COSMIC RAYS AND THE EARTH’S ATMOSPHERE
A.D. Erlykin1,2 and A.W. Wolfendale2 1 P.N. Lebedev Physics Institute, Leninsky Prospekt, Moscow, Russia. 2 Department of Physics, University of Durham, South Road, Durham, DH1 3LE, U.K.
Abstract Avery brief summary is given of aspects of cosmic ray physics which have rel- evance to the possible effects of cosmic rays on ‘climate’. It is concluded that a more detailed look at the effect of fast ionizing particles on the atmosphere from the standpoint of cloud production would be advantageous.
1. INTRODUCTION There have been many claims for a correlation between solar properties (e.g. sunspot number) and climate but all have suffered from the absence of a reasonable physical cause, the point being that the energy changes in the solar variations are deemed to be too small to account for the necessary climate forcing. This is not to say that there are no well-documented effects on very long time scale; there are. Those due to variations in the sun-earth distance and the inclination of the earth’s axis (Milankovic effects), which relate to 103 − 106 y periods, are generally agreed. What is not (yet) agreed is that the 11-year solar cycle has a significant correlation with climate. The best evidence favouring a specific cause for a sunspot (SS) — climate connection relates to the apparent role of cosmic rays which are, themselves, modulated by solar activity. The observation is that by the Danish group in which there is a correlation of cloud cover and CR intensity over the oceans. The likelihood of this being genuine comes from the fact that CR are the major source of ionization away from land and CR, of course, provide ionization. Insofar as the cloud cover/CR intensity results are considered in detail elsewhere, more discussion will not be given here; rather, we will concentrate on the CR aspects.
2. COSMIC RAY INTENSITY AS A FUNCTION OF ATMOSPHERIC DEPTH It is relevant to consider the manner in which the vertical intensity of cosmic rays varies with height in the atmosphere. Considering the three major components: protons, electrons and muons, the values of −2 −1 −1 IV , the vertical intensity in cm s sr ,atheights above sea level of 2, 5 and 10 km, respectively, are:
− − − p :2× 10 4, 1.6 × 10 3 and 2 × 10 2 − − − e :5× 10 3, 3 × 10 2 and 2 × 10 1 − − − µ :1× 10 2, 2 × 10 2 and 5.5 × 10 2.
Some comments can be made, as follows: (i) The peak of the ionization (for µ and e)isinthe 10 km region, much higher than the common cloud level. Such an observation is not ‘the kiss of death’ to the correlation idea, because of uncertainty in transport phenomena for the products of CR ionization between the 10 km level and much lower levels. (ii) Although the ionization produced by protons is lower than that produced by e it is important to point out that the rare ‘stars’, produced by proton interactions in the atmosphere, contain very highly ionizing nuclear fragments. It is not inconceivable that subtle effects leading to cloud droplets are associated with these highly ionizing fragments. 3. KNOWN METEOROLOGICAL EFFECTS Concerning the CR intensity at ground level there are three major ‘meteorological variations’: (i) The pressure coefficient, due simply to absorption of the secondaries, of −2% per cm Hg. (ii) A correlation with the height of the 100 mb level, amounting to ∼−5% per km. The reason is to do with µ − e decay. (iii) The mean air temperature between the 100 and 200 mb levels. The dependence is +0.1% per ◦K and is due to π − µ decay. All the effects are well understood by the cosmic ray fraternity. However, they should be borne in mind by the wider community when correlations are sought.
4. COSMIC RAY ORIGIN — AND EFFECTS 4.1 Galactic particles Below the ‘knee’ in the energy spectrum, at ∼ 3 × 1015 eV, it is probable that most CR come from supernova remnants (SNR) by way of shock acceleration. It is these particles, mainly — specifically, below about 1012 eV — that are modulated by the solar wind which has, itself, an 11-year cycle. It is inevitable that there should be intensity variations due to the stochastic nature of SNR but these should be rare. There has been a claim for a 2-fold increase in intensity some 35 thousand years ago but it seems likely (Beer, private communication) that the increase is due not to an SNR but to a variation of the earth’s magnetic field and/or solar variability.
4.2 Solar particles With the advent of space vehicles, the study of the (mainly) low energy solar CR has become a ‘growth industry’. The range-energy relation is such that most protons, or their progeny, do not reach ground level; even at 10 GeV the particles only reach a height of about 10 km. Nevertheless, solar CR are important, particularly in the polar regions where effects on the ozone layer have been claimed, and their presence is eminently reasonable. Finally, we can remark on the possibility of very rare solar flares having serious effects on climate and, indeed, mankind itself. An extrapolation of the log N − log S curve for energy deposited on earth above S would indicate very serious effects every million years, or so. However, this interval is surely too short (otherwise we would not have survived for so long!). What can be said is that significant effects might be expected every 1000 years, or so. Effects on climate might occur if the atmosphere happened to be in an unstable phase at the time.
5. CONCLUSIONS Cosmic ray effects offer the possibility of being relevant to climate change. Although it is premature to be dogmatic, the likelihood of significant climatic effects is high enough for a detailed analysis of the physics — and meteorology — of CR-air interactions to be not just desirable, but vital.
ACKNOWLEDGEMENTS The authors are grateful to Jasper Kirkby for re-kindling our interest in this topic. ICE CORE DATA ON CLIMATE AND COSMIC RAY CHANGES
J. Beer Federal Institute of Environmental Science and Technology, EAWAG, CH-8600 Dübendorf, Switzerland, tel: +41 1 823 51 11 / fax: +41 1 823 52 10, email: [email protected]
Abstract Ice cores represent archives which contain unique information about a large variety of environmental parameters. Climatic information is stored in the form of stable isotopes, greenhouse gases and various chemical substances. The content of cosmogenic nuclides such as 10Be and 36Cl provide long-term records of the intensity of the cosmic ray flux and its modulation by solar activity and the geomagnetic dipole field. Cosmogenic nuclides are produced by the interaction of cosmic ray particles with the atmosphere. After production, these nuclides are transported and distributed within the environment, depending on their geochemical properties. Some of them are removed from the atmosphere by snow and incorporated into ice sheets and glaciers. The analysis of the Greenland ice cores GRIP and GISP2 are discussed in terms of climate and cosmic ray changes during the past 50’000 years.
1. ARCHIVE ICE Polar ice sheets are formed from snow. The snowflakes grow together to grains which slowly increase in size. Due to the pressure of the overlying new snow layers, the grains become more and more compacted and finally turn into ice. The consequence of this formation process is that the ice not only preserves all the atmospheric constituents such as aerosols and dust, it also contains air bubbles that enable to determine the atmospheric composition and in particular the reconstruction of greenhouse gases in the past. This unique property makes ice the only archive that virtually stores all the climate forcing factors (greenhouse gases, aerosols and volcanic dust, solar irradiance) except internal variability. Ice cores also contain information on the corresponding climate response (temperature, precipitation rate, wind speed, atmospheric circulation). Another important property of ice is that it flows. This can be seen in Fig. 1, which schematically depicts an ice-sheet. The ice slowly flows towards the margin of the ice sheet, where it partly melts and partly breaks up as icebergs. Under steady-state conditions, the ice lost in the ablation area is replaced by snow falling on the accumulation area where new layers are formed continuously. As a consequence of the horizontal movement of the ice, the annual layers become thinner with increasing depth, as indicated in Fig. 1. This leads to another special property of the archive ice. The depthÐage relationship is non- linear, which has the advantage that the uppermost part of the core is well resolved and the total time period covered is long (of the order of 105 years for polar ice cores). The disadvantage of this non-linear time-scale is, however, that dating ice is difficult and relies strongly on correct modeling of the ice-flow. The main ice sheets are situated in polar regions (Greenland, with a maximum thickness of approx. 3 km and Antarctica, with a thickness of up to 4 km). Smaller ice sheets at lower latitudes can only be found at high altitudes (Andes, Himalayas, Alps) [1]. There is a steadily growing number of parameters which can be measured in ice cores. It is beyond the scope of this paper to discuss all these parameters. In Table 1, a small selection of those related to climate forcing and climate response is given. ❄❄❄❄❄❄❄❄ ❄ ❄ ❄❄❄ ❄ ❄❄❄
ACCUMULATION
ABLATION ABLATION
BEDROCK
Figure 1: Formation of an ice sheet. The snow falling in the accumulation region turns into ice that slowly flows towards the ablation area where it breaks up into ice-bergs or melts. As a consequence of the flow characteristics the thickness of annual layers decreases with increasing depth.
Table 1. Climate parameters measured in ice cores.
Parameter Proxy for
CO2 Greenhouse gases
CH4 Greenhouse gases
SO4 Volcanic eruptions Ash Volcanic eruptions 10Be, 36Cl Solar activity 18 d O Temperature Borehole temperature Temperature Annual layer thickness Precipitation rate Dust Wind speed Anions / cations Atmospheric circulation
18 18 As an example, d O of the GRIP ice core is shown in Fig. 2. d O (relative deviation of the 18O/16O ratio in ice from a standard in ‰) reflects mainly the temperature at which snow is formed.
-30
-35
18 d O [‰]
-40
-45 01020304050 Age [ky BP]
18 Figure 2: d O measured in the GRIP ice core from Greenland. Low values indicate cold climate. During the last 10'000 years the temperature was relatively stable compared to the preceding glacial period. Figure 2 shows that during glacial times the temperature in Greenland was characterized by abrupt changes (so-called Dansgaard-Oeschger events) of up to 20∞ C within a few decades. The last 10’000 years, the so-called Holocene, however, looks comparatively stable. The Dansgaard- Oeschger events were probably caused by abrupt changes in the ocean circulation, transporting heat to high latitudes. In the following, we will concentrate on what cosmogenic radionuclides in ice cores can tell us.
2. COSMOGENIC RADIONUCLIDES IN ICE The cosmic ray particles (87% protons, 12% helium nuclides, 1% heavier particles) that enter the Earth’s atmosphere react with Nitrogen, Oxygen and Argon, producing a cascade of secondary particles. These nuclear processes produce a variety of cosmogenic nuclides such as 10Be, 14C and 36Cl. These nuclides are listed in Table 2 together with their main properties.
Table 2. Some cosmogenic radionuclides and their main properties.
Nuclide Half-life Target Production rate (years) (atoms cm-2 s-1) 10Be 1.5 106 N, O 0.018 14C 5730 N, O 2.0 36Cl 3.01 105 Ar 0.0019 The physics of the production processes is well understood and therefore the production rate can be calculated for each point in the atmosphere, depending on the heliospheric modulation and the geomagnetic field intensity, provided the involved nuclear cross-sections are known [2]. As an example, Fig. 3 shows the dependence of the mean global production rate of 10Be as a function of solar modulation parameter F (F = 0: quiet sun, F = 1’000: very active sun) and the geomagnetic field intensity B in relative units (B = 1 corresponds to the present field intensity). As can be seen, the dynamic range between no magnetic field (B = 0), no solar modulation (F = 0) and doubled magnetic field (B = 2), very active sun (F = 1’000) is about one order of magnitude. Note that the dependencies are non-linear and that production changes by only a factor 3-4 were observed so far.
4
3
2
Be Production Rate 1 10
Rel. 0 0.5 Geomagnetic Field 1 0 200 1.5 400 600 800 2 1000 F) Solar activity (
Figure 3: Dependence of the relative mean global 10Be production rate on the geomagnetic field intensity and the solar activity parameter F. The production rate 1 corresponds to a geomagnetic field 1 and a F of 550 corresponding to the average solar activity. The transport of the cosmogenic nuclides produced in the atmosphere is not as well 14 understood as the production processes. C forms CO2 and exchanges between the main reservoirs of the carbon cycle (atmosphere, ocean, biosphere). 10Be and 36Cl become attached to aerosols or exist in gaseous form (H36Cl). After a mean residence time of 1 to 2 years they are removed from the atmosphere mainly by wet precipitation. In Polar Regions, the aerosols are removed by the snow that forms the ice sheet. Assuming a production rate of 0.018 10Be atoms cm-2 s-1 (Table 2) and a precipitation rate of 100 cm y-1, a simple calculation reveals an average 10Be concentration of approximately 107 atoms per kg of ice. Extremely sensitive detection techniques are necessary to measure 107 atoms. Due to the long half- life, decay counting is not feasible. However, accelerator mass spectrometry (AMS), using single atom detection is suitable to do the job [3]. A known amount of stable 9Be (typically 0.5 mg) is added to each sample. This leads to a 10Be/9Be ratio in the range of 10-13 to 10-12. After extraction of the Be from the water by ion exchange technique, a BeO sample is produced. This sample is put into the ion source of the AMS system and an ion beam is produced and accelerated to high energy (20 MeV) by means of a tandem accelerator. This high energy destroys the molecular background and enables suppression of the isobaric background (10B in the case of 10Be). In the following, some of the results obtained so far are discussed:
3. GEOMAGNETIC MODULATION To reconstruct the geomagnetic field from the 10Be and 36Cl fluxes we assume that the 10Be and 36Cl fluxes at Summit are proportional to their average global production rate.
geomagneticK field intensity (deduced from Be-10 and Cl-36 data) 1.5 geomagnetic field intensity (Mediterranean Sea)
1
0.5 Geomagnetic field intensity [relative to the present level]
0 20 30 40 50 60 Age BP [kyrs]
Figure 4: Comparison between the geomagnetic field reconstructed from a combined 10Be - 36Cl record from the GRIP ice core [4] with the paleomagntic data derived from a mediterranean sediment core [5].
Figure 4 shows the geomagnetic field intensity for the period 20-60 kyr BP, reconstructed from the combined 10Be and 36Cl flux in the GRIP ice core [4]. The shaded area indicates the uncertainty in the calculated field intensity. Also shown is the field measured on a Mediterranean sediment core [5]. The correlation between the geomagnetic field intensities obtained from these two independent reconstructions is very high (r2=70%). In our calculation of the geomagnetic field intensity, the combined 10Be and 36Cl flux was normalized in such a way that a value of 10% of its current value is assumed for the minimum of the calculated geomagnetic field intensity at about 40 kyr BP (Laschamp event). The normalization is also supported by new data from sediment cores of the Atlantic ocean [6]. The Laschamp event corresponds to a period of increased cosmic ray flux and therefore provides a test case for the proposed relationship between cosmic ray flux, cloud cover, and climate change [20]. The maximum of the combined flux of 10Be and 36Cl should be correlated 18 with the d O data (note the inverse scale) and CH4 data (Fig. 5). However, this is clearly not the case. During the Laschamp event (36-41.5 kyr B.P.) the combined flux of 10Be and 36Cl is not 18 2 2 significantly (p < 0.1) correlated with either d O (r = 0.07%) or CH4 (r = 0.09%). The same applies over the entire time interval shown in Figure 5 (r2 = 0.3% and 0.4%, respectively) [7]. A 2 (a) (10Be + 36Cl) (proxy for 1.8 cosmic ray flux)
Cl)-flux 1.6 36 1.4 (rel. units) Be + 1.2 10 ( 1
(b) B 18O -42 d (proxy for climate) -41
-40
O [‰] -39
18 corr.(A,B): d -38 r2=0.2%
(c) 350 C CH4 (proxy for climate) 400
450 [ppbv] 4 500 corr.(A,C): CH 550 r2=0.4%
20 25 30 35 40 45 50 55 60
Age BP [kyr]
10 36 18 Figure 5: Comparison of the combined Be- Cl flux with the climate parameters d O and CH4. According to the proposed relationship between cosmic ray flux and climate (Svensmark, this volume) a correlation between the three parameters is expected for the Laschamp geomagnetic minimum (shaded area) which is not present [7].
4. SOLAR MODULATION Direct observations clearly reveal that part of the solar variability is cyclic. In the following, we will concentrate only on cycles with time scales of years and longer. Cycles with periodicities from centuries to millennia are based on indirect or proxy data. Since these data (e.g. 10Be, 14C) represent a complex combination of different signals it is not always possible to unambiguously attribute a cycle to solar variability. One way of distinguishing between solar variability induced signals and others is to compare 10Be and 14C. Both radionuclides are produced by similar nuclear reactions in the atmosphere. Their respective production rate and their dependence on solar activity can be calculated [2]. However, after production their geochemical behaviour differs completely. A comparison of the two radionuclide records therefore allows us to distinguish between the production signal caused by solar and geomagnetic modulation and the system signal caused by the climate affecting the transport and the exchange processes between the different reservoirs. The results from such comparisons indicate that, for the past several millennia, the short- 14 term (decades to centuries) fluctuations in the D C record are mainly due to production variations, most probably caused by solar modulation. It is important to note that cycles associated with solar activity do not have a fixed periodicity. For example in the case of the sunspot cycle, the periodicity varies between 9 and 17 years. This raises the important question whether the periodicity averaged over longer times remains constant or not [8, 9]. To answer this question, longer and very precisely dated records of solar activity are needed than are presently available. The most prominent solar cycle is the 11-y Schwabe cycle discovered by Schwabe in 1843 when analysing his 17 year-long sunspot data. In Fig. 6, the sunspot cycle based on sunspot groups [10] is shown for the period 1600-1999 together with the inversely plotted 10Be concentration measured in the Dye 3 ice core from South Greenland [9].
0.5 10 Be [10 4 g -1 ] 1
100 Dalton Maunder 50 Sunspots
0 1600 1650 1700 1750 1800 1850 1900 1950 2000 YEAR
Figure 6: Comparison of sunspot numbers with 10Be concentration. Periods of local reduced solar activity are dashed.
In view of the fact that sunspot numbers and heliospheric modulation of the 10Be production rate are different representations of a common cause, i.e. solar activity, the agreement is good. A detailed analysis shows that the 10Be signal lags behind the sunspot signal by about 1 year, corresponding to the mean residence of 10Be in the atmosphere. It is interesting to note that the Schwabe cycle is still present in the 10Be record during the Maunder minimum [11]. A 90-year cycle was discussed by Gleissberg when analysing the auroral record [12]. The Dye 3 annual 10Be record going back to 1423 also shows the 90-year Gleissberg cycle [13].
14 The 205-year DeVries cycle is the most prominent periodicity in the D C record during the Holocene. However, as with other periodicities, its amplitude and periodicity are variable with time. Since the sunspot record is too short to detect the 205-year DeVries cycle, its attribution to solar variability is based on indirect evidence. Cycles with longer periodicities (e.g. 1000-2000 years) could not yet be attributed to solar modulation. An especially interesting feature of the sunspot record is the period from 1645 to 1715 A.D. which is characterized by an almost complete absence of sunspots (Fig. 6), the so-called Maunder minimum. Since then, solar activity has steadily grown with the exceptions of a few less pronounced minima: the Dalton minimum (1790-1830) and some weaker minima around 1890 and 1960. 40
30
20
10 C [‰]
14 0 D
-10
-20 7000 6000 5000 4000 3000 2000 1000 0 Year BP
Figure 7: 14C peaks corresponding to periods of low solar activity and possibly also reduced solar irradiance
Maunder type minima occurred earlier throughout the Holocene and are called grand minima. Since only little is known about these grand minima from direct observations, their occurrence is documented mainly by cosmogenic nuclide records. Fig. 7 shows the detrended 14 D C record [14]: The grand minima that correspond to maxima with regard to the cosmogenic 14 nuclide production are marked with arrows. How do we know that these peaks in D C are of solar origin and not caused by climatic effects or geomagnetic modulation? Firstly, the similarity in amplitude and duration of the peaks with the one corresponding to the Maunder minimum points to a common cause. Secondly, it seems rather unlikely that the geomagnetic dipole field would exhibit such strong excursions within only approximately a century. Finally, the good agreement 14 14 10 between the measured D C and the calculated D C based on Be data from ice cores convincingly shows that these peaks are due to production and not climatic effects. This brings us to the last topic, solar forcing and its detection [15].
5. SOLAR FORCING OF CLIMATE CHANGE The two main problems related to solar forcing and climate change are: 1. The lack of a quantitative solar forcing function. The physical processes responsible for changes in solar irradiance are not yet well understood, especially as far as long-term changes are concerned. All attempts so far are therefore based mainly on various assumptions leading to differences of about a factor of 2. Longer forcing records are based on simple linear regression models [16]. There may be other effects on the atmosphere caused by the interaction of the heliosphere with the magnetosphere and by cosmic rays with the atmosphere which could also contribute to climate change [17]. 2. The response function of the climate system to solar forcing is probably variable in time and not well known. There is an increasing number of experiments with global circulation models (GCM) including solar forcing. However, these model runs do not take into account the change in the spectral energy distribution and its potential effects on the atmosphere (e.g. ozone). In view of all these uncertainties, which would be the best strategy to detect solar induced climate changes? One approach is to use the Milankovic forcing that is caused by planetary gravitational effects on the orbital parameters of the earth [18]. Although only changes of the eccentricity causes changes in the total solar irradiance, the fact that the latitudinal forcing function can be calculated precisely for any time offers the unique opportunity to study the response function on longer time-scales (≥ 10 ky). Another straightforward approach is to search for fingerprints of solar forcing. All we know for sure is that solar irradiance changed in phase with solar activity over the past two Schwabe cycles. It is reasonable to assume that longer solar activity changes are associated with larger changes in solar irradiance [15]. Therefore, good candidates for solar forcing effects are solar minima, in particular grand minima. In fact, instrumental temperature records reveal cold events during the local minima around 1810, 1890 and 1960 (Fig. 5). The Maunder and Spoerer minima occurred during the so-called “little ice-age”, a period characterized by a general advance of glaciers. The more high-resolution climate records become available, the more evidence is found that abrupt climate changes indeed often coincide with solar minima (Van Geel, this volume)[19]. With regard to the question of the underlying physical mechanisms of solar forcing, a crucial test is the phase relationship. While the proposed mechanism of cosmic ray induced cloud formation tolerates no phase shift between cosmic ray flux and climate response [20], this is less the case for changes in solar irradiance that may be coupled by slow processes with the modulation of cosmic rays.
6. CONCLUSIONS Ice cores contain a large number of proxies for different climate parameters such as for 18 temperature (d O), greenhouse gases (CO2, CH4) and aerosols (chemical constituents). In the form of cosmogenic nuclides (10Be, 36Cl) they also provide unique information about the cosmic ray flux which is modulated by the geomagnetic dipole field and the solar activity that can be traced back in time over the past ca. 60'000 years. The suggested relationship between geomagnetic field, galactic cosmic rays, and climate could not be confirmed for the period of the Laschamp event (36-41.5 kyr B.P.). 10Be measurements show that solar variability has a cyclic component with periodicities of 11, 90, 205 and possibly more years. However, the relationship between solar activity and solar irradiance is not yet understood in detail.
ACKNOWLEGMENTS The author thanks W. Mende, R. Muscheler and G. Wagner for helpful discussions and C. Wedema for typing the manuscript and improving the English. This work was financially supported by the Swiss National Science Foundation.
REFERENCES [1] Thompson, L.G., et al., Tropical climate instability: the Last Glacial cycle from a Qinghai- Tibetan ice core. Science, 276: p. 1821-1825 (1997) [2] Masarik, J. and J. Beer, Simulation of particle fluxes and cosmogenic nuclide production in the Earth's atmosphere. J. Geophys. Res., 104(D10): p. 12,099 - 13,012. (1999) [3] Suter, M., et al., Advances in AMS at Zurich. Nucl. Instrum. Meth., B40/41: p. 734-740. (1989) [4] Wagner, G., et al., Reconstruction of the geomagnetic field between 20 and 60 ka BP from cosmogenic radionuclides in the GRIP ice core. Nucl. Instr. Meth., B 172: p. 597-604.( 2000) [5] Tric, E., et al., Paleointensity of the geomagnetic field during the last 80,000 years. J. Geophys. Res., 97: p. 9337-9351.( 1992) [6] Laj, C., et al., North Atlantic paleointensity stack since 75 ka (NAPIS-75) and the duration of the Laschamp event. Phil. Trans. R. Soc. Lond., 358: p. 1009-1025 (2000) [7] Wagner, G., et al., Some results relevant to the discussion of a possible link between cosmic rays and the Earth’s climate. J. Geophys. Res., 106(D4): p. 3381-3388.( 2001) [8] Dicke, R.H., Is there a chronometer hidden deep in the Sun? Nature, 276: p. 676-680.( 1978) [9] Beer, J., et al., Solar Variability Traced by Cosmogenic Isotopes, in The Sun as a Variable Star: Solar and Stellar Irradiance Variations, J.M. Pap, et al., Cambridge University Press: Cambridge. p. 291-300.Editors. (1994) [10] Hoyt, D.V. and K.H. Schatten, Group sunspot numbers: a new solar activity reconstruction. Solar Physics, 179: p. 189-219.( 1998) [11] Beer, J., S.M. Tobias, and N.O. Weiss, An active Sun throughout the Maunder minimum. Solar Physics, 181: p. 237-249.( 1998) [12] Gleissberg, W., The eighty-year solar cycle in auroral frequency numbers. J. Br. Astron. Assoc., 75: p. 227.( 1965) [13] Beer, J., et al., 10Be as an indicator of solar variability and climate, in The solar engine and its influence on terrestrial atmosphere and climate, E. Nesme-Ribes, Springer-Verlag: Berlin. p. 221-233. Editor. (1994) [14] Stuiver, M., et al., INTCAL98 Radiocarbon age calibration, 24,000-0 cal BP. Radiocarbon, 40(3): p. 1041-1083.( 1998) [15] Beer, J., W. Mende, and R. Stellmacher, The role of the Sun in climate forcing. Quat. Sci. Rev., 19(1-5): p. 403-415.( 2000) [16] Bard, E., et al., Solar irradiance during the last 1200 years based on cosmogenic nuclides. Tellus, 52B: p. 985-992.( 2000) [17] Tinsley, B.A., G.M. Brown, and P.H. Scherrer, Solar Variability Influences on Weather and Climate: Possible Connections Through Cosmic Ray Fluxes and Storm Intensification. J. Geophys. Res., 94(D12): p. 14783-14792.( 1989) [18] Berger, A., et al., eds. Milankovitch and Climate. NATO ASI Series. Vol. 126, D. Reidel Publishing Company: Dordrecht. Vol 1 und 2 zusammen: 895.( 1984) [19] Magny, M., Solar influences on Holocene climatic changes illustrated by correlations between past lake-level fluctuations and the atmospheric 14C record. Quaternary Research, 40: p. 1-9.( 1993) [20] Svensmark, H. and E. Friis-Christensen, Variation of cosmic ray flux and global cloud coverage - a missing link in solar-climate relationships. J. Atm. Terr. Phys., 59(11): p. 1225-1232.( 1997) LONG-TERM VARIATIONS IN COSMIC RAY FLUXES AND TOTAL SOLAR IRRADIANCE AND THEIR POSSIBLE INFLUENCE ON GLOBAL CLIMATE CHANGE
M. Lockwood *) Space Science and Technology Department, Rutherford Appleton Laboratory, Oxfordshire, UK
Abstract Studies of how geomagnetic activity is excited by the solar wind flow have allowed quantification of the open magnetic flux of the Sun, revealing it to have more than doubled during the 20th century. This flux fills the heliosphere out to the termination shock and shields Earth from galactic cosmic rays: thus, were air ions produced by cosmic rays to facilitate the formation of clouds in any way, this magnetic field would modulate terrestrial cloud cover. We here confirm there is a strong and statistically significant anticorrelation between the heliospheric field and the global coverage of low-altitude (<3.2 km) clouds and discuss the implications for extrapolating cloud-cover estimates back in time. We also show that the correlation between clouds and cosmic rays and the anticorrelation between clouds and total solar irradiance (TSI) are very similar in their strength and significance, making distinction between potential TSI and cosmic ray effects difficult to achieve.
1. INTRODUCTION The aa index of geomagnetic activity was devised by Mayaud in 1972 and, on annual timescales at least, successfully quantifies global geomagnetic activity from just two, antipodal observatories [1]. The importance of this index lies in the fact that it is a homogenous data series that extends back to 1868. Lockwood et al. have recently used the aa data to infer long-term changes in the open flux of the Sun that threads the coronal source surface and is dragged into the heliosphere by the solar wind flow [2]. The method they devised was an inversion of the analysis of Stamper et al. [3] who used data from solar cycles 20, 21 and 22, for which regular spacecraft measurements of the near-Earth heliosphere are available. The method was refined by Lockwood and Stamper [4] who used only cycles 21 and 22 to determine the required coefficients and held back the heliospheric field measurements from cycle 20 as an independent test of the method. The RMS differences between the inferred and observed radial components of the heliospheric field Br for the test cycle 20 were actually smaller than for the fitted cycles 21 and 22. Further confirmation of the method comes from the very high and highly significant anticorrelation between the inferred open solar flux and the counts detected by various neutron monitors due to cosmic ray bombardment of the atmosphere [5]: 80% of the variation of the cosmic ray flux could be associated with the heliospheric field strength. A number of different processes contribute to the shielding of cosmic rays [6], but scattering by irregularities in the heliospheric field is a dominant effect [7], such that the shielding is dependent on the open solar flux. Included in the remaining 20% that is not explained by the variation in open solar flux, is the known effect of solar cycle number on cosmic ray fluxes at Earth. This is expected theoretically as a consequence of the gradient and curvature drifts associated with large-scale heliospheric structure [6]. The effect reverses with the polarity of the polar solar field, roughly 1 year after the
*) Also at Department of Physics and Astronomy, Southampton University, Southampton, UK. peak of each cycle, and is apparent at the data, predominantly at solar minimum when the heliospheric field is weakest [7, 8, 9]. The method used to derive the heliospheric field is based on the theory of solar wind Ð magnetosphere coupling by Vasyluinas et al. [10], and thus by extrapolating back in time to before cycle 21 we are assuming that no there is no additional unknown factor, not included in this theory, the behaviour of which is different on decadal and century timescales. Given the correlation obtained for cycles 21 and 22 is 0.97, to be relevant this factor would need to have introduced variability before the start of cycle 21, but not subsequently. An important point in this respect is that the correlation between the open solar flux inferred from geomagnetic activity and cosmic ray neutron products was equally high and significant for solar cycles 19, 20, 21 and 22. Thus the method has been confirmed by independent data from cycles 19 and 20 Ð despite the fact that cycle 19 was the largest solar activity cycle ever observed and that cycle 20 was surprisingly weak. The key finding from the method is that the heliospheric field, averaged over full solar cycles, increased by 140% over the 20th century [2]. Extrapolating using the correlations with the cosmic ray fluxes discussed above, yields that the flux of primary cosmic rays was some 15% higher on average in 1900 than at present for energies above 3 GeV and 4% higher for >13GeV [5]. Support for this inferred drift comes from the abundance of the 10Be found, for example, in the Dye-3 Greenland ice core [11, 12]. This is formed as a spallation product when cosmic rays impact O and N in the atmosphere, and is then deposited in the ice sheet by precipitation. The dependence of precipitation on climatic conditions introduces scatter, nevertheless a clear anticorrelation with the inferred open solar flux is found [5]. In addition, the variation of 14C found, for example, in tree rings is consistent with the change seen in 10Be [13]. The complication for both these isotopes is that the abundances detected are subject to climate change. However, the effect is very different in the two cases: 10Be is precipitated into ice sheets, a process that introduces lag and a dependence on climate, whereas 14C is directly absorbed in gaseous state but has reservoirs in the biomass and oceans, exchange with which masks the true cosmogenic production rate and is expected to vary with climate. However, the similarity of the inferred century-scale changes in 10Be and 14C production rates strongly implies that the cause is a variation in cosmic rays and not climatic. Svensmark and Friis-Christensen [14], Svensmark [15] and Marsh and Svensmark [16] have discussed a correlation between cosmic ray fluxes and global cloud cover on Earth. The correlation is best with higher energy cosmic ray fluxes and low-altitude cloud cover. The present paper contains three studies. Given that the open solar flux quantifies 80% of the cosmic ray variation, section 2 looks at the direct correlation between open solar flux and cloud cover. We then use the long-term variation in the open flux derived from the aa index to look at the possible change in global cloud cover since 1900, assuming the correlation were real and not influenced by any other factors. One possibility for such a factor is the total solar irradiance (TSI) of the Sun, which is now known to show a solar cycle variation [17, 18, 19] and which also shows an upward drift over the 20th century in a variety of reconstructions that employ proxy data [4, 20, 21, 22]. The open solar flux, for the interval of the global cloud cover data at least, is well correlated with the TSI [4, 23]. This correlation was originally found in annual mean data and before observations for the rising phase of solar cycle 23 became available [4]. However, recent work [23] has shown that the correlation, although somewhat lower in monthly averages (correlation coefficient, r = 0.61), is highly significant (>99.99%), and has been maintained in solar cycle 23. Section 3 compares the correlations between cosmic rays and cloud cover and TSI and cloud cover. Section 3 also considers the effect of temporal smoothing on the significance of these correlations. 2. CLOUD COVER AND OPEN SOLAR FLUX Figure 1 shows the time series of monthly means for 1984-1994 of the strength of the heliospheric field near Earth (|B|), the aa index and the percentage change in global cloud cover from the average,
Figure 1. Monthly means of : (a) the magnitude of the heliospheric field near Earth ( |B| , blue line) ; (b) the aa index (green and black line); and (c) -
2.1 Correlation Analysis The correlogram for annual means of
It can be seen that a good match to ccf is again achieved. The peak anticorrelation is at a lag of Ð1 month, more consistent with a solar wind propagation delay. The peak is r = Ð0.53, a weaker correlation coefficient than in Figure 2, but a statistically much more significant result because there is much less “persistence” or “conservation” in the unsmoothed data. We can quantify the significance using the Student’s t-test, by making a correction to the number of degrees of freedom to allow for persistence [30]:
1/2 Ne = N (1 - a1)/(1 + a1) , t = |r| {(Ne - 2)/(1 - r)} , (1) where N is the number of samples, Ne is the effective number of samples, and a1 is the mean autocorrelation at lag 1 of the input and the output. The acf at lag 1 for
0.40 for monthly data, giving a1 = 0.505, the t statistic derived from (1) then yields a significance -4 of 99.98% (i.e. the probability that this correlation is a chance occurrence is just 2¥10 ). 2.2 Extrapolation back in Time Data on the IMF magnitude |B| only extends back to 1963 [26]. However, we can go further back in time if we use the coronal source flux, Fs, which has been estimated from the aa index [2] and is related to |B| by: 2 2 Fs = (1/2). 4pR1 |Br| = 2pR1 |B| cos( g ) (2) where R1 is 1AU (the mean Earth-Sun distance), Br is the radial component of the IMF and g is the IMF garden hose angle at Earth [2, 23]: on annual timescales g is almost constant [2, 3] and thus |B| and Fs are approximately linearly related.
Figure 3. Same as Figure 2, but for monthly mean data
The uncertainties inherent in the use of equation (2) have recently been analysed in detail and are of order ±20% in monthly data and ±10% in annual data [23]. The implications for the past behaviour of global cloud cover depend on whether the relationship implied by the above correlation is linear or not. Figure 4 shows the scatter plot for the annual means of
Thus the data strongly imply that global cloud cover was higher around 1900 than it is now. However, before we can say this with certainty and quantify the factor involved, we need to understand the physical and chemical mechanisms of the interaction and so we can understand the regression fit and know which is the most appropriate functional form to use.
3. CLOUD COVER AND TOTAL SOLAR IRRADIANCE The correlation of global, low-altitude cloud cover is significantly higher for cosmic rays than for the 10.7 cm radio flux from the Sun [14, 15, 16]. However, at this wavelength, the solar emission is far from representative of the IR, optical, and UV emissions that dominate energy input into the terrestrial climate system (rather, F10.7 is very closely related to sunspot activity and is most relevant to Earth’s upper atmosphere, the thermosphere). A sequence of total solar irradiance (TSI) values covering more than 2 solar cycles has been compiled by Fröhlich and co-workers [18, 19]. This compilation requires careful intercalibration of the various space-based radiometers used and allowance for their degradations with time and exposure. The data reveal a solar cycle variation with TSI being of order 0.1% greater at sunspot maximum than at sunspot minimum. Figures 7 and 8 show the scatter plots of the integrated, global, low-altitude cloud cover
Figure 5. Extrapolated low-altitude global cloud cover estimates for 1868-1998: (black) from a linear fit to observation; and (blue) from a square-law loss. The observed data are shown in red.
Figure 6. Detail from Figure 5 for 1980-2000. Figure 7. Scatter plot of monthly means of global cloud cover
Figure 8. Scatter plot of monthly means of
Figure 9 shows the time series of monthly means for the
Figure 10. Same as figure 7, for 12-point running means of the monthly data
Figures 10, 11 and 12 correspond to 7, 8 and 9, respectively, but are for 12-point running means of the monthly data. It can be seen that the long-term variations are very similar in the data and the scatter in the scatter plots has been greatly reduced. For
Figure 11. Same as figure 8, for 12-point running means of the monthly data
4. CONCLUSIONS AND DISCUSSION The fraction of the globe covered by low-altitude clouds has shown a solar cycle variation in recent data [14, 15, 16]. With little more than one cycle of heterogeneous data, we cannot be sure that this is truly an oscillation that will continue to match the solar cycle variations so well. The temporal correlation between the low-altitude global cloud cover and cosmic ray counts is highly significant in monthly data. Introducing smoothing increases the correlation coefficient magnitude and makes the time-series plots appear to be in greater agreement, but also removes the statistical significance from the correlation. An anti-correlation between the total solar irradiance (TSI) and global cloud cover is found to have the same strength and significance as the correlation with cosmic ray fluxes. Thus we cannot tell if it is more likely to be the cosmic rays or the TSI that are influencing cloud cover. In addition, because of the nature of all correlation studies, cannot we be sure that either TSI or cosmic rays have a causal effect on cloud cover. A parametric study using a global coupled ocean-atmosphere model is required to see if we would expect this anticorrelation between TSI and low-altitude cloud cover. Whether caused by TSI or comic ray variations, the presence of a solar cycle signal in global cloud cover would effectively be an amplification of the solar effect on Earth’s climate. Recent modelling using the UK Hadley Centre’s HAD3CM global coupled ocean-atmosphere model has pointed towards the presence of such an amplification, both when fitting the 11-year solar cycle variation in the amplitude of the global spatial pattern of average tropospheric temperatures, and when fitting the 150-year drift in the global average surface temperature [M.R. Allen et al., Private communication, 2000]. In both cases an amplification factor of about 2.5 was needed to gain the best fit, compared to the value from radiative forcing arguments. To within the 90% confidence level, this factor varied between 1 and 6.
Figure 12. Same as figure 8, for 12-point running means of the monthly data
Somewhat surprisingly, this amplification of the solar influence calls for an amplification factor for the man-made influences that go into the model [M.R. Allen, private communication, 1999]. This amplification factor is much smaller than for the solar effect (of order 1.1). The key reason for this behaviour is that the main drift in the longer-term TSI variation took place between 1900-1950. This also happened to coincide with a period of reduced volcanic activity (a global cooling phenomenon). On the other hand, anthropogenic greenhouse gasses have had their dominant effect in the past 30 years. Underestimating the solar effect early in the 20th century effectively causes the model to fit an anthropogenic effect that starts earlier but is less steep.
ACKNOWLEDGEMENTS The author thanks Nigel Marsh and Henrik Svensmark for the provision of the D2 cloud cover data and the World Data Centre system for the cosmic ray, heliospheric and geomagnetic data. He is also grateful to Myles Allen, Nigel Marsh, Henrik Svensmark, Jasper Kirkby, and many other scientists for valuable discussions. This work was funded by the UK Particle Physics and Astronomy Research Council.
REFERENCES [1] P.N. Mayaud, J. Geophys. Res., 77, 6870-6874, 1972. [2] M. Lockwood et al., Nature, 399, 437-439, 1999. [3] R. Stamper, et al., J. Geophys Res., 104, 28,325-28,342, 1999 [4] M Lockwood and R. Stamper, Geophys Res. Lett., 26, 2461-2464, 1999. [5] M. Lockwood, J. Geophys Res., 106, 16021-16038, 2001 [6] J.R. Jokipii in “The Sun in Time”, eds. C.P. Sonnet, M.S. Giampapa and M.S. Matthews, Univ. of Arizona Press, pp. 205-221, 1991 [7] H.V. Cane, Geophys. Res. Lett., 26, 565-568, 1999 [8] I.G. Usoskin, J. Geophys.Res., 103, 9567-9574, 1998. [9] H.S. Ahluwalia, J. Geophys. Res., 102, 24,229-24,236, 1997 [10] V.M. Vasyliunas, Planet Space Sci., 30, 359-365, 1982. [11] K.G. McCracken and F.B. McDonald, The long-term modulation of the galactic cosmic radiation, 1500-2000, in press, in Proc. 27th. Int. Cosmic Ray Conference, Hamburg, 2001 [12] J. Beer et al., Sol. Phys., 181, 237-249, 1998. [13] E. Bard, et al., Earth and Planet. Sci. Lett., 150, 453-462, 1997. [14] H. Svensmark, and E. Friis-Christensen, J. Atmos. Sol. Terr. Phys., 59, 1225-1232, 1997. [15] H. Svensmark, Phys. Rev. Lett., 81, 5027-5030, 1998. [16] N. Marsh, and H. Svensmark, Space Sci. Rev., 94, (1/2), 215-230, 2000. [17] R.C. Willson, Science, 277, 1963-1965, 1997. [18] C. Fröhlich, and J. Lean, Geophys. Res. Lett., 25, 4377-4380, 1998. [19] C. Fröhlich, Space Sci. Rev., 94, (1/2), 15-24, 2000. [20] J. Lean, et al., Geophys. Res. Lett., 22, 3195-3198, 1995. [21] S.K. Solanki and M. Fligge, Geophys. Res. Lett., 26, 2465-2468, 1999. [22] Hoyt, D., and K. Schatten, J. Geophys. Res., 98, 18,895-18,906, 1993. [23] M. Lockwood, An evaluation of the correlation between open solar flux and total solar irradiance, Astron and Astrophys., in press, 2001. [24] Y.-M. Wang, Geophys. Res. Lett., 27, 621-624, 2000. [25] W.B. Rossow, et al., International Satellite Cloud Climatology Project (ISCCP): Documentation of new datasets, WMO/TD 737, World Meteorol. Organ., Geneva, 1996. [26] D. A. Couzens and J. H. King, Interplanetary Medium Data Book - Supplement 3, National Space Science Data Center, Goddard Space Flight Center, Greenbelt, Maryland, USA, 1986. [27] M. Lockwood, The relationship between the near-Earth Interplanetary field and the coronal source flux: dependence on timescale, J. Geophys. Res., in press, 2001. [28] M.S. Potgieter, Adv. in Space Res, 16(9), 191-203, 1995. [29] A.C. Cummings et al., J. Geophys. Res., 99, 11,547-11,552, 1994. [30] Wilks, D.S., Statistical methods in the atmospheric sciences, Academic Press, San Diego, California, USA, 1995. EVIDENCE FROM THE PAST: SOLAR FORCING OF CLIMATE CHANGE BY WAY OF COSMIC RAYS AND/OR BY SOLAR UV?
Bas van Geel1, Hans Renssen2 and Johannes van der Plicht3 1 Institute for Biodiversity and Ecosystem Dynamics, Universiteit van Amsterdam, Kruislaan 318, 1098 SM Amsterdam, The Netherlands [email protected] 2 Institut d«Astronomie et de Géophysique G. Lemaître, Université Catholique de Louvain, 2 Chemin du Cyclotron, B-1348 Louvain-la-Neuve, Belgium [email protected] 3 Centre for Isotope Research, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands [email protected]
Abstract Major Holocene shifts to cool and wet climate types in the temperate zones correspond to suddenly increasing values of the atmospheric 14C content, suggesting a link between changing solar activity and climate change. In the temperate zones the transition from the Subboreal to the Subatlantic (ca 850 cal BC) represents a sudden, strong shift from a relatively dry and warm climate to a humid and cool episode. The moment of change occurred at, or maybe even just before the start of a sharp rise of the atmospheric 14C content. In previous studies, we postulated two amplification mechanisms: a) increased cosmic ray flux causes an increase in atmospheric 14C content, and also a climate shift, b) a decline of solar UV causes a reduced stratospheric ozone concentration, leading to climate change at the earth surface. Two phenomena indicate that mechanism a) is much less likely than mechanism b): 1) The enhancement of cosmic ray intensity to relatively high levels took place several decades after the climate shift. 2) In Central Africa and in Western India there was a shift to dryness. Chronological differentiation in solar output may play a role, but this is purely hypothetical.
1. INTRODUCTION Over the last few hundred years, changes in solar irradiance have been relatively small (less than 1 W/m2). As a consequence, solar forcing of abrupt climate change has been controversial [1]. However, there is strong evidence from the past for an important role of the sun upon climate change [2-5]. To explain this past evidence of solar forcing, we postulated two possible amplifying mechanisms that could explain how relatively small changes in solar irradiance could lead to abrupt climatic shifts [6]. a) Changes of cosmic ray intensity (modulated by fluctuating solar wind) might have an effect on cloud formation and thus on the planetary albedo and on temperature [7], and/or b) Within the small changes of solar activity, changes in UV are important [1]. Changes in solar UV have an effect on ozone formation in the lower stratosphere. Variations in the ozone concentration modulate the stratospheric temperature, leading to changes in the stratospheric circulation that could be propagated downwards to the Earth’s surface, thus influencing atmospheric circulation patterns world-wide [8, 9]. We review the evidence for solar forcing of climate change at the Subboreal-Subatlantic transition, as found in raised bogs and other paleodata, and we evaluate the possible contribution of both mechanisms mentioned above.
2. RAISED BOG AS ARCHIVES OF PAST CLIMATE Peat deposits are valuable archives for paleoclimate studies. The so-called raised bogs in NW- Europe are rainwater fed and the paleohydrological changes of such bogs mainly reflect climate shifts. A climate shift around 850 calendar years BC is visible in raised bog profiles as a transition from peat which was formed during a period of a relatively warm climate (darker, more decomposed peat) to lighter coloured upper peat, formed during a period of cooler, wetter climatic conditions. We use the Radiocarbon (14C) method for precise dating of climate-induced transitions in peat layers. Radiocarbon ages are expressed in BP, Radiocarbon "years" relative to 1950 AD. Radiocarbon years are different from calendar years because the production of 14C has not been constant in the past due to changes in both the geomagnetic field strength and in solar activity. The 14C time scale is calibrated by measuring the 14C content of tree rings, dated absolutely by means of dendrochronology [10]. The solar activity changes characterise the calibration curve by means of fluctuations (the so-called "wiggles"). Calibration of a single Radiocarbon date usually yields an irregular probability distribution in calendar age, quite often over a long time interval. This is problematic in paleoclimatological studies, especially when a precise temporal comparison between different climate proxies is required. However, a sequence of (uncalibrated) 14C dates can be matched to the wiggles in the calibration curve (wiggle-match dating [11, 12]). A high-resolution 14C sample sequence can result in a precise chronology of the peat core. This dating strategy also revealed relationships between atmospheric 14C variations and short-term climatic fluctuations (as detected in peat deposits) caused by solar variations. Data from Holocene lake deposits in the Jura Mountains also strongly point to a relationship between 14C fluctuations and paleohydrological shifts under the influence of climate change [2]. The climate shift around 850 cal BC (Subboreal-Subatlantic transition) was one of the most important climate shifts during the Holocene. We focused on this transition, which was immediately followed by a sharp rise of the atmospheric 14C content during the period between 850 and 760 cal BC. We identified the peat-forming mosses (representatives of the genus Sphagnum) in peat profiles of Northwest and Central European raised bogs. Knowing the ecological preferences of the mosses, we could interpret the recorded changes in species composition in terms of hydrological changes, related to climate change [13, 14]. Before the climate shift from the Subboreal to the Subatlantic period, Sphagna of the section Acutifolia were important peat formers in the Dutch bogs. Then Sphagnum papillosum and Sphagnum imbricatum took over. This change of the peat-forming plants indicates a shift from relatively warm, to cooler, wetter climatic conditions. The paleo-record from raised bogs shows that the abrupt climate shift happened at, or maybe even shortly before, the start of the period of the sharply rising atmospheric 14C content (Figure 1). In various lowland regions in the Netherlands where settlement sites were present, the climate shift at the Subboreal-Subatlantic transition caused a considerable rise of the ground water table so that arable land was transformed into wetland, where peat growth started. Bronze Age farmers living in such areas had to migrate because they could no longer produce enough food in their original settlement areas. Like the raised bog evidence, the archaeological evidence also points to a climate shift just preceding the enhanced cosmic ray intensity [13]. We also found strong evidence for climate change around 850 cal BC in other parts of the world [15, 16 and references therein]. In the temperate zones of Europe, North America and South America there is evidence for an equatorward shift of suddenly enhanced Westerlies, while the climate changed (cooling, higher effective precipitation). Fig. 1: The radiocarbon calibration curve (lower diagram) for the period between 1000 to 500 BC and 14 corresponding atmospheric fluctuations (D C, upper diagram). The moment of the climate shift, which precedes the rise of the atmospheric radiocarbon content, is indicated with an arrow. 3. THE CONTRIBUTION OF AMPLIFYING MECHANISMS The observed climate changes around 850 cal BC may have been caused by the lowering of solar irradiation through two amplifying factors, namely, (1) increased cosmic ray intensity stimulating
Polar Cell Polar Easterlies PFJ 60°N polar front Prevailing Ferrel Westerlies Cell 30°N STJ
N.E. Trades Hadley Cell
0° ITCZ
S.E. Trades
° 30 S STJ Prevailing Westerlies
60°S Polar PFJ Easterlies
Fig. 2A Simplified model of the tropospheric circulation (similar to the present situation) before the discussed climate change around 850 cal BC. ITCZ: Intertropical Convergence Zone; STJ: Subtropical Jet; PFT: Polar Front Jet.
Polar Polar Easterlies Cell 60°N polar front PFJ Prevailing Westerlies Ferrel Cell 30 N ° STJ N.E. Trades Hadley Cell 0 ° ITCZ
S.E. Trades STJ 30°S Prevailing Westerlies PFJ 60 S ° Polar Easterlies
Fig. 2B: As in Fig. 2A, but for the period directly after the climate change around 850 cal BC. Grey arrows denote changes which may be summarised as follows: equatorward shift of location of Jets, expansion of polar cells (i.e., cooling in mid-latitudes), relocation of mid-latitude storm tracks (regional increase in precipitation), and reduction of strength of Hadley Cells (i.e., drier conditions in the tropics). cloud formation and possibly also precipitation in certain regions, and (2) reduced solar UV intensity, causing a decline of stratospheric ozone production and cooling as a result of less absorption of sunlight. Figure 2 [after Ref. 6] shows the effect a decline of solar UV would have on the atmospheric circulation near the Earth’s surface [compare Refs. 8 and 9]: a decrease in the latitudinal extent of Hadley Cell circulation (weakening of the monsoon) may have occurred with concomitant equatorward relocation of mid-latitude storm tracks [see also Ref. 17]. This picture fits in with the paleoclimatological evidence from the northern and southern temperate zones (cooler, wetter) and the contemporaneous dryness crisis in Central Africa and Western India, which is evident from pollen records and archaeological evidence [13, 16]. The evidence during the Subboreal-Subatlantic transition strongly supports the "Haigh model" [8, 9] as an effective amplification mechanism for changes in solar activity. The combination of detailed paleoclimatological data from different parts of the world delivers circumstantial evidence for the suggestion that the UV-ozone mechanism had more effect on climate than the mechanism related to the increase of the cosmic ray intensity. In summary, we conclude that there is paleo-evidence for solar forcing of climate change around 850 cal BC. Of the two possible amplification mechanisms, the reduced UV scenario was most likely the effective one. There seem to be two arguments against an important role of cosmic rays (cloud formation) in relation to climate change: 1) Detailed series of radiocarbon dates from archaeological sites and raised bogs [11, 13, 14, 18] show that the abrupt climate change around 850 cal BC had already occurred when the cosmogenic isotope 14C only started to show an initially insignificant rise. This is supported by data for 10Be (another cosmogenic isotope, the production of which more directly reflects changing cosmic ray intensities than 14C), showing a corresponding and more or less contemporaneous rise as shown by 14C. For this event there might be a delay of approximately 10 years in the 14C rise only [J. Beer, pers. comm.; compare Ref. 19]. Consequently, the time- lag in the rise of the 14C content (compared to climate change) cannot be attributed to possible delaying processes related to the carbon cycle. In other words: the strong rise in cosmic ray intensity only followed climate change, and thus cannot have triggered the change [compare Ref. 12 for major climate shifts during the Little Ice Age in relation to similar increases of atmospheric Radiocarbon]. 2) The widespread dryness in the tropics (weaker monsoon in Central Africa and Western India) after 850 cal BC is not an effect that is expected to occur with enhanced cloud formation under the influence of increased cosmic ray intensity. However, the dryness in the tropics may not be inconsistent, as climatic teleconnections are not always straightforward (e.g., in the case of El Niño) and it could be that cooling in the mid-latitudes (where enhanced cloud formation due to increased cosmic ray intensities may be favoured) has resulted in drying in some regions in the tropics. On the other hand, it must be noted that the observed world-wide, but strongly contrasting changes in climate at the Subboreal-Subatlantic transition fit remarkably well in the model for an important role of solar UV (see Fig. 2). An important role for the reduced UV-scenario would raise one, yet unanswered, question: could a considerable decline of solar activity indeed have chronologically different phases (effective electromagnetic signal before magnetic signal; so first a UV decline with strong effects on climate, and later a more gradual decline of solar wind affecting the increased production of cosmogenic isotopes)? Solar physicists might be able to answer this question. Alternatively, detailed observations of variations in solar activity in the near future may reveal a solution to the question about which amplification mechanism plays a role in solar forcing of climate change.
ACKNOWLEDGEMENTS We thank Jürg Beer and Raimund Muscheler for critical reading of the manuscript and Dmitri Mauquoy for correction of the text. REFERENCES [1] D.V. Hoyt and K.H. Schatten, The role of the sun in climate change, Oxford, 1997 (Oxford University Press) [2] M. Magny, Quat. Res. 40 (1993) 1. [3] B. van Geel et al., Quat. Sc. Rev. 18 (1999) 331. [4] D.A. Hodell et al., Science 292 (2001) 1367. [5] U. Neff et al., Nature 411 (2001) 291. [6] B. van Geel and H. Renssen, In: Water, Environment and Society in Times of Climatic Change (Kluwer, Dordrecht, 1998) p. 21-41. [7] H. Svensmark and E. Friis-Christensen, J. Atm. Sol. Terr. Phys. 59 (1997) 1225. [8] J.D. Haigh, Nature 370 (1994) 544. [9] J.D. Haigh, Science 272 (1996) 981. [10] M. Stuiver et al., Radiocarbon 40 (1998) 1041. [11] M.R. Kilian et al., , 2000. Quat. Sc. Rev. 19 (2000) 1011. [12] D. Mauquoy et al., Evidence from North-West European bogs showing that Little Ice Age climatic changes were driven by changes in solar activity. Holocene 12: in press. [13] B. van Geel et al., Radiocarbon 40 (1998) 535. [14] A. Speranza et al., Quat. Sc. Rev. 19 (2000) 1589. [15] B. van Geel et al., 2000. Holocene 10 (2000) 659. [16] B. van Geel et al., 2001. In: Y. Yasuda and V. Shinde (Eds), Monsoon and Civilization, Extended Abstracts of the 2nd International Workshop of the Asian Lake Drilling Programme (Pune, India, 2001) p. 35. [17] D. Shindell et al., Science 284 (1999) 305. [18] B. van Geel et al., J. Quat. Sc. 11 (1996) 451. [19] R. Muscheler et al., Terra Nostra 3 (2001) 156. THE ROLE OF CLOUD COVER VARIATIONS ON THE SOLAR 13 ILLUMINATION SIGNAL RECORDED BY d C OF A SHALLOW WATER IONIAN SEA CORE (1147-1975 AD)
G.Cini Castagnoli, G. Bonino, D.Cane and C.Taricco Dipartimento di Fisica Generale dell'Università, Via P.Giuria 1, 10125 Torino, and Istituto di Cosmogeofisica del CNR, Corso Fiume 4, 10133 Torino, Italy
Abstract 13 We show the d C profile of Globigerinoides ruber, measured in the GT90/3 shallow-water Ionian sea core, dated with high accuracy (better than 1%) using radiometric and tephroanalysis methods,. It is commonly 13 accepted that d C variations in symbiotic foraminifera mainly record the effects of symbiont density and of photosynthetic activity, varying with ambient light level. The core, extracted from the Gallipoli platform, was sampled at contiguous steps of thickness 2.5 mm, corresponding to 3.87 13 years. The d C profile covers the period 1147-1975 AD. During the first seven centuries it appears fairly flat, while it shows a steep increase between 1760 and 1950 of ~0.3‰. The analysis of the time series performed using different methods shows a dominant decadal periodicity throughout the record. The 11-year component is identified at high significance level by Monte Carlo singular spectrum analysis (MC-SSA); the SSA-reconstructed-11-year component is in phase with the sunspot
solar cycle. The average amplitude of this component is A11y=0.04‰. 13 The modern d C increase (induced by a light level increase) of about 0.3 ‰ is concomitant with the decrease of the number of cloudy days per 13 year of about 11% at the site of the core deposition. If also the d C 11-y cycle has its origin in the modulation of cloudiness, the observed variation of 0.08 ‰ (peak-to-trough) requires an 11 y-cloud cover cycle (paced by the sun) of about 11%*0.08‰/0.3‰=3%. This is of the same order of the 11 y solar cloudiness cycle proposed by Svensmark and Friis-Christensen for the recent solar cycles, on a global scale (1980- 1995).
1. INTRODUCTION The carbon isotopic ratio 13C/12C in the shells of symbiont-bearing foraminifera is controlled by symbiont density and by their photosynthetic activity [1], i.e. by the primary productivity of the habitat. Provided the isotopic ratio of the bath is known at the time of the shell growth [2, 3], the isotopic ratio can be utilised for the quantitative study of paleoceanic and paleoatmospheric processes. Since Urey proposal in 1947 [4], many isotopic measurements have been performed for elucidating the climatic changes in the geological past. But only few stable-isotope time series covering the last millennium are available, which may be used to determine recent-past changes. This happens in spite of their importance for understanding the evolution of the present-day environmental conditions. A few carbon stable isotope time series were studied in different archives, mainly corals (see e.g. Ref. [5]), but they cover only the last few centuries. Sediments with a high sedimentation rate, which allows high resolution, may offer the opportunity to study in detail the millennial time scale; however it is difficult to find a suitable site for an absolute dating. We have found the right characteristics in the Gallipoli Terrace (Gulf of Taranto, Ionian sea) at a water depth of about 200 m. The carbonatic sediment is deposited at a sedimentation rate constant over the last 2 millennia [6-8].
13 In this paper we present the d C signal in specimens of the symbiotic planktonic foraminifera Globigerinoides ruber of the shallow water Ionian sea core GT90/3.
13 The d C time series, covering the last 828 years with the time resolution of 3.87 years, gives us the possibility to acquire information a) on the modulation of light at sea surface by the solar irradiance and by the cloud coverage in this region over the past millennium; b) on the presence of an 11-year signal most likely forced by the solar cycle and c) on the importance of this interdecadal variation of solar origin with respect to the variations of the trend.
2. THE IONIAN SEDIMENTS
The coast and the whole Salentina Peninsula are very flat and there is no direct river discharge on the platform where we took the cores. We have extracted many cores in different coring campaigns. We performed an accurate dating by radiometric and tephroanalysis methods [6,9]. The sedimentation in the cores shows no obvious laminae or discontinuities; dating is based upon 210 226 evaluation of Pb (T1/2 = 22.3 years) "excess" with respect to the activity supported in situ by Ra 210 222 (T1/2=1600 years). The "excess" Pb is atmospheric fallout from decay of Rn (T 1/2 =3.82 days). Core dating by this method is restricted to ages not greater than 150 years. The high correlation coefficient between the profile of "excess" 210Pb in the sediment and a decreasing exponential provides information on the constancy of the sedimentation rate over the past two centuries. Checks on the Pb age profile and on its extrapolation to the whole core are obtained from a 137Cs spike at 1963-1964 AD, due to a peak in nuclear weapons testing, on the one hand, and from tephroanalysis, on the other. The latter identifies clinopyroxene sedimentation peaks corresponding to the well-known historical volcanic eruptions at Pompei (79 AD), Pollena (472 AD), Ischia (1301 AD), Monte Nuovo (1538 AD) and starting from 1631 AD up to the present identifies the minor peaks corresponding to the detailed registration of the volcanic activity of the Vesuvius by the Vesuvian Observatory [10]. The position of the Gallipoli Terrace is particularly favourable for the collection, in the core mud, of the volcanic markers, fallout of the Campanian area activity, because the westerly winds bring the ashes towards the Gulf of Taranto. The sedimentation rate s was found to be quite constant along the cores and uniform throughout the -1 whole platform in the last millennia: we determined s = (0.0645±0.0007) cm year ; therefore the core depth scale may be transformed in a time scale, accurate better than 1%. The volcanic markers allow also to infer that bioturbation is not effective in the region at least within the adopted sampling interval of 0.25 cm, corresponding to 3.87 years. In fact the number density of pyroxenes in the volcaniclastic layers, with sharp boundaries, is identical in different cores taken in the area (see Fig.2 in Ref.[6]). The presence of 137Cs at the proper core-depth guarantees that the top of the core has not been perturbed. In Fig.1 we show the carbonate profiles of the cores GT14, GT89/3 and GT90/3 (all from the same area), sampled at contiguous steps of the same thickness 0.25 cm to determine the total carbonate content (CaCO3). We may notice the remarkable correlation between the carbonate records of different cores, demonstrating the uniformity of the deposition of the platform. In the same figure, we present (at the base) the number density pyroxene record measured in the upper 130 cm of the cores. It provides exact time "benchmarks" starting from the first historical eruption of the Vesuvius, described by Plinius, which destroyed the cities of Pompei and Ercolano in 79 AD. We notice that we have found large peaks only in correspondence to the volcanic events historically recorded. In these sediments we have studied the profiles of different bulk properties of the mud, with the primary aim of providing time series useful for investigating solar-terrestrial relationships in the past millennia (see e.g. Ref.[11]). Recently, we have chosen to measure the stable isotope composition of G. ruber planktonic foraminiferal tests [12-14]. G. ruber, a surface-warm water- dwelling foraminifer, shows maximum abundance in the top 20 m of the mixed layer in early autumn when the thermocline begins to break down [15]. This symbiont bearing foraminifer and 13 13 therefore the d C of its shell, like the d C of other species [16], is mainly controlled by the symbiont photosynthetic activity and by the ambient irradiance levels.
sample number 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 100 0 45 40 35 GT89-3 GT14
30 45 25 40 35 45 30 40
35 GT90-3 30 Ischia 1301 A.D. 25 200 Pompei Pollena Montenuovo 79A.D. 472 A.D. 1538 A.D. 100
0 0 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 500 1000 1500 year (AD)
Fig.1. Carbonate profiles (percentage of CaCO3 in the sediment’s mud) of the three shallow-water Ionian sea cores GT14, GT89/3, GT90/3. We may read the CaCO3 concentration as a function of sample number (upper scale) and as a function of time (lower scale). The reference (top) level is 1979 AD. In the lower part of the figure, the pyroxenes profile is also plotted, clearly showing that the principal peaks of the last 2 millennia are those caused by the eruptions of Pompei, Pollena, Ischia and Montenuovo.
3. EXPERIMENTAL PROCEDURE
We sampled the GT90/3 core (39∞45'53"N, 17∞53'33"E, water depth 174 m) at 0.25 cm thickness intervals, from the top down to sample 215, in a continuous sequence, covering the time interval 1147 AD-1979 AD. Samples of about 5 g of sediment were soaked in 5% calgon solution over night then treated in 10% H2O2 to remove any residual organic material, subsequently washed with distilled water jet through a sieve (150 mm). The fraction > 150 mm was kept and oven-dried at 50∞C. G-ruber were picked up under microscope. For each sample, 20-30 specimens of G.ruber of the same size were selected for the isotopic measurements, which were performed using a VG- PRISM mass spectrometer fitted with an automated ISOCARB preparation device. Analytical precision based on internal standards was better than 0.1‰ [13]. Calibration of the mass Spectrometer to VPDB scale was done using NBS19 and NBS18 carbonate standards.
13 4. THE d C TIME SERIES AND ITS ANALYSIS 13 In Fig.2 we show the d C profile (mean value xm = 0.84‰, s = 0.14‰), consisting of a continuous record of N = 215 points from 1147 to 1975; the sampling interval is ts = 3.87 years. 1.3
1.1
0.9
0.7
0.5
0.3 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 years (AD)
13 Fig.2 d C profile measured in Globigerinoides Ruber of the shallow water Ionian sea core GT90/3 (red line). This profile covers the period 1147-1975 AD, with a resolution of 3.87 years.
The 5 points running average (heavy-black line) and the SSA-reconstructed trend from PCs 1-2-5 are superposed to the data. The most evident feature of the series is the rapid enrichment in 13 d C starting from about 1760. In order to obtain reliable results, the analysis of this time series was performed using different spectral methods, like periodogram, correlogram, maximum entropy method, superposition of epochs (SE) and singular spectrum analysis (SSA; see, e.g., the review paper of Ref.[17]; here we present the results obtained by the classical Blackman-Tukey (BT) correlogram ([18]; see also Ref.[19]) and by SSA [20-22]; the results are also tested using a Monte Carlo approach (MC-SSA) [23,24]. The typical problems of the classical spectral estimates (power leakage and high variance) has therefore to be tackle for the correlogram. In order to reduce the effect of power leakage (due to the implicit window of the time series) and to give a consistent estimate of the true spectrum, a Bartlett (triangular) window was applied to the autocorrelation function. We observe that different window (like, for example, the Hamming or Hanning window) give very similar spectral results. The variance reduction is obtained using a window length M 0.1 0.09 ~400 y 0.08 0.07 0.06 0.05 11.3 y 0.04 0.03 0.02 0.01 0 0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (year-1) 13 Fig.3. Blackman-Tukey power spectrum of the d C time series. A Bartlett window (width 50) was applied to the autocorrelation function. The FFT spectrum was computed using 512 frequencies. Note the prominent peak at 11.3 years and the power at the low frequencies of the trend. 10 9 8 7 6 5 11 y (EOFs3-4) 4 3 2 1 0 010203040506070 eigenvalue order k Fig.4. SSA-variance spectrum, plotted as percent of the total variance (normalized eigenvalues lk) associated to each of the 65 eigenvectors EOFs of the CD matrix, in decreasing order of variance. EOFs 3-4, associated with the 11.3 years oscillation, carries ~9% of the total variance. The trend is represented by EOFs 1-2-5 (~20% of the total variance). In order to perform a good signal-to-noise separation, we use the Monte Carlo method (MC-SSA) [23,24]. In this approach, we assume a model for our time series (null-hypothesis) and we determine the parameters using a maximum-likelihood criterion. Then a Monte Carlo ensemble of surrogate series is generated from the model and SSA is applied to data and surrogates (EOFs of the null-hypothesis basis are used), in order to test whether it is possible to distinguish the series from the ensemble. Since a large class of geophysical processes generates series with larger power at lower frequencies, we have assumed AR(1) noise in evaluating evidence for trend and oscillations. This is done to avoid overestimation of the system predictability, having underestimated the amplitude of the stochastic component of the time series [24]. In our case, we adopt the size 10000 for the Monte Carlo ensembles and we assume at first a pure noise AR(1) null-hypothesis: in this case we note anomalous power at frequencies corresponding to periods of ª11 years and to the trend; this is significant at 98% level. We confidently reject this hypothesis since the eigenvalues corresponding to the 11-year oscillation and to the trend component stand above the Monte Carlo range; moreover the 11-year range is of interest a priori and we expect a low frequency enhancement due to the presence of an evident trend in the series. Now we assume AR(1)+trend (EOFs 1-2-5) null-hypothesis. The result of the test against this hypothesis is shown in Figure 5.1, where we have plotted the eigenvalues and the surrogate bars as a function of the dominant frequency associated with the corresponding EOFs of the composite null-hypothesis basis. The vertical bars indicate the range in which lie the 98% of the eigenvalues determined from the ensemble of Monte Carlo simulations. We note that EOFs with period of ª11 years still show more variance than expected on this null-hypothesis. This is significant at 98% level. Finally, we test against AR(1)+trend+11years (EOFs 3-4) null-hypothesis; since in Figure 5.2 there are no excursions above the 98% MC-bars, we cannot reject this hypothesis. 1 5.1) 99th & 1st percentiles 0.1 11 years trend 0.01 null-hypothesis AR(1)+EOFs 1-2-5 (trend) 0.001 0 0.1 0.2 0.3 0.4 0.5 1 5.2) 99th & 1st percentiles 0.1 11 years trend 0.01 null-hypothesis AR(1)+EOFs 1-2-5 (trend) + 3-4 (11 years) 0.001 0 0.1 0.2 0.3 0.4 0.5 frq. assoc. with EOF-k (cycles in 3.87 years) 13 Fig.5 Application of Monte Carlo SSA to the d C time series (the empty squares show the eigenvalues associated with the EOFs included in the null-hypothesis). The Monte Carlo ensemble size is 10000. 5.1) Test 13 of d C series against the composite null-hypothesis of AR(1) noise plus trend. EOFs corresponding to the 11.3- year oscillation show exceeding power with respect to this hypothesis. This is significant at the 98% level. 5.2) 13 Test of d C series against the AR(1) noise plus 11.3-year oscillation and trend. No excursions occur outside the 98% limits, than we cannot reject this null-hypothesis and therefore the series can be explained by this model. 13 Therefore, our spectral analysis suggests that the d C time series is composed of a trend, on which an oscillation of about 11 years is superimposed; the background consists of AR(1) noise. Applying the method of SE to the reconstructed components (principal components PCs 3- 4), we determine the average amplitude A11y=0.04‰ of the interdecadal signal over the 828 years covered by the series. 5. DISCUSSION 13 The connection of the d C-11-year component to the solar cycle may be inferred by inspection 13 of Fig.6, where the reconstructed d C component (from PCs 3-4) is compared with the sunspot number series from 1700 to 1975. 200 ~11 y cycle (RCs 3,4) 0.1 Sunspots 150 100 0.0 50 -0.1 0 1750 1800 1850 1900 1950 2000 year (AD) Fig.6. Comparison of the SSA-reconstructed component (~ 11.3-years, from PCs 3-4), heavy line, with the 13 sunspot number series, light line, from 1750 to 1975. The variations of d C and of sunspot number are in 13 phase, in such a way high solar activity corresponds to high d C values. 13 The two signals are in phase, suggesting a solar forcing of d C, in such a way that high solar 13 output corresponds to high d C values. This favours the point of view that higher solar irradiance, as seen in space at sunspot maximum in the last 20 years [25-26], induces higher primary productivity of the symbionts (algae living on the spines of the G.ruber), which, using preferentially 12C for growing their tissue, leave carbon enriched of 13C in the chamber of G.ruber, for the construction of the shell. The primary productivity could in principle be modified also by other factors (nutrients availability, etc); however changes in illumination seem to be the most effective on carbon isotope fractionation processes: in this case, one of the important sources of 13 d C variations, beside irradiance solar variability, are the changes from year to year of the cloud coverage. The cloudiness seems to be paced by the solar cycle through galactic cosmic ray (GCR) modulation, as suggested by Svensmark and Friis-Christensen [27] and Svensmark [28]. Higher solar activity corresponds to lower GCR flux, giving lower cloud coverage, thus reinforcing the effect of the enhanced irradiation. Unfortunately a quantitative evaluation of all the above mechanisms involved in the formation of the 13C signal in G.ruber is not yet available. A decrease in cloud coverage has been observed from 1875 to 1975 in the region in which we took the core. The meteorologist De Giorgi has collected an accurate homogeneous archive of rainfall data from 1875 to 1921 in Lecce (Puglia). In the paper of Mangia et al. [29] those data have been integrated up to 1980 by using data taken in Bari. A decrease of annual rainy days (with rainfall >0.2 cm) of about 11% between 1875 and 1980 has been reported. In figure 7 we 13 compare the d C data (dashed line) with the number of rainy days per year (solid line). 1.3 40 50 1.2 60 1.1 70 80 1 90 0.9 100 13C - GT90/3 d d d d 0.8 110 120 rainy days per year 0.7 130 0.6 140 1860 1880 1900 1920 1940 1960 1980 year (AD) 13 Fig.7. d C modern increase in GT90/3 core (dashed line) and decrease in rainy days per year (inverted scale, solid line) from 1875 to 1980. We note that the rainy days decrease of 40 days per year, i.e. of ~11%, is concomitant with 13 the d C modern increase of ~0.3‰. The Apulian observations are confirmed by other evidences. Buffoni et al. [30] have analysed the precipitation in Italy from 1833 to 1996. On a yearly basis a decreasing trend in precipitation of the order of 10% has been found statistically significant. Furthermore Russo et al. [31] have observed in the long running time series of rainy days over Genova a decrease of about 10% between 1833 and 1992 with a negative tendency throughout the whole period. Therefore the experimental decrease in our area of about 11% of rainy days between 1875 13 and 1980 can be used for interpreting the carbon isotope effect. If the d C modern increase of 13 0.3‰ is due to change in illumination produced by a cloud cover decrease of 11% and if the d C 11y cycle of 0.08‰ (peak-to-trough) is attributed to the same process, we deduce a cloud cover variation over the 11-y cycle of 0.08‰*11%/0.3‰=3%. This value is in agreement with the variation of the global cloud coverage, paced by the solar cycle, between 1980 and 1995, as given by Svensmark and Friis-Chrstensen [27] and Svensmark [28]. ACKNOWLEDGEMENTS We are grateful to Prof. Carlo Castagnoli for his support and discussions and to Alberto Romero for continuous technical assistance. This work was supported by MURST-Co-fin 98, 2000 and by CNR. REFERENCES [1] Spero, H.J. and Williams, D.F., 1988, Extracting environmental information from 13 planktonic foraminiferal d C data, Nature 335, 717-719. [2] Spero, H.J., Lerche, J. & Williams, D.F., 1991, Opening the carbon isotope "vital effect" black box, 2: Quantitative model for interpreting foraminiferal carbon isotope data. Paleoceanography 6, 639-655. [3] Spero, H.J., Bijma, J., Lea, D.W. and Bemis, B.E., 1997, Effect of seawater carbonate concentration on foraminiferal carbon and oxygen isotopes Nature 390, 497-500. [4] Urey, J.C., 1947, J.Chem.Soc. 562. [5] Boiseau, M,, Ghil, M., and Juillet-Leclerc, A., 1999, Climatic trends and interdecadal variability from South-Central Pacific coral records, Geophys.Res.Lett. 26, 2281-2884. [6] Cini Castagnoli, G., Bonino, G., Caprioglio, F., Provenzale, A., Serio, M. & Zhu, G.M., 1990, The carbonate profile of two recent Ionian sea cores: evidence that the sedimentation rate is constant over the last Millennia Geophys.Res.Lett. 17, 1937-1940. [7] Cini Castagnoli, G., Bonino, Provenzale, A., Serio, M. & Callegari, E., 1992a, The CaCO3 profiles of deep and shallow Mediterranean sea cores as indicators of past solar-terrestrial relationships, Nuovo Cimento C15, 547-563. [8] Cini Castagnoli, G., Bonino, G., Serio, M. & Sonett, C.P., 1992b, Common spectral features in the 5500-year record of total carbonate in sea sediments and Radiocarbon in tree rings. Radiocarbon 34, 798-805. [9] Bonino, G., Cini Castagnoli, G., Callegari E., and Zhu, G.M., 1993, Radiometric and tephroanalysis dating of Recent Ionian Sea cores, Nuovo Cimento C 16, 155-161. [10] Arno’, V., Principe, C., Rosi, M., Santacroce, R., Sbrana, A. and Sheridan M.F., 1987, Eruptive History, in Summa-Vesuvius, Quaderni de La Ricerca Scientifica, CNR, 114, (ed. Santacroce R.) 53-103. [11] Cini Castagnoli, G., 1997, The 200y variability in terrrestrial archives: solar forcing? in Past and Present variability of the Solar-Terrestrial System: measurements, data analysis and theoretical models (eds. Cini Castagnoli, G. and Provenzale, A. 47-57, I0S press, Amsterdam). [12] Cini Castagnoli, G., Bernasconi, S.M., Bonino, G., Della Monica, P., Taricco, C., 1999a, 700 year record of the 11 year solar cycle by planktonic foraminifera of a shallow water Mediterranean core, Adv. Space Res. 24, 233-236. [13] Cini Castagnoli, G., Bernasconi, S.M., Bonino, G., Della Monica, P., and Taricco, C., 18 1999b, Solar activity in the last millennium recorded in the d O profile of planktonic foraminifera of a shallow water Ionian sea core, Sol. Phys. 188, 191-202. [14] Cini Castagnoli, G., Bonino, G., Taricco, C., 2001, Long term solar-terrestrial records from sediments: carbon isotopes in planktonic foraminifera in the last millennium, Adv. Space Res., in press. [15] Beveridge, N.A.S., and Shackleton, N.J., 1994, Carbon isotopes in recent planktonic foraminifera: a record of anthropogenic CO2 invasion of the surface ocean, Earth and Planet. Sci. Lett. 126, 259-273. [16] Spero, H.J. and Lea, D.W., 1993, Intraspecific stable isotope variability in the planktonic foraminifera Globigerinoides sacculifer: Results from laboratory experiments. Mar. Micropaleontol. 22, 221-234.0 [17] Ghil, M., and Taricco, C., 1997, Advanced spectral-analysis methods, in Past and Present Variability of the Solar-terrestrial System: Measurement, Data Analysis and Theoretical Models Proc. of the Int. School of Physics E.Fermi, Varenna, 1996, Course CXXXIII, edited by G.Cini Castagnoli and A.Provenzale (IOS Press, Amsterdam), p.137-159. [18] Blackman, R.B., and Tukey, J.W., 1958, The Measurement of Power Spectra from the Point of View of Communication Engineering (Dover, New York). [19] Kay, S.M., 1988, Modern Spectral Estimation: Theory and Applications (Prentice-Hall, Cliffs, N.J.). [20] Broomhead, D.S., and King, G.P., 1986, Extracting qualitative dynamics from experimental data, Physica D 20, 217-236. [21] Vautard, R., and Ghil, M., 1989, Singular spectrum analysis in nonlinear dynamics, with applications to paleoclimatic time series, Physica D 35, 395-424. [22] Dettinger, M.D., Ghil, M., Strong, C.M., Weibel, W., and Yiou, P., 1995, Software expedites singular-spectrum analysis of noisy time serie, EOS Trans. AGU, 76, 12. [23] Allen, M.R., 1992, Interactions between the atmosphere and oceans on timescales of weeks to years, PhD Thesis, Clarendon Laboratory, Oxford. [24] Allen, M.R., and Smith, L.A., 1996, Monte Carlo SSA: detecting irregular oscillations in the presence of coloured noise, J.Clim.9, 3373. [25] Willson, R.C., and Hudson, H.S., 1991, The Sun’s luminosity over a complete solar cycle, Nature 351, 42-44. [26] Pap, J.M., 1997, in Past and Present Variability of the Solar-terrestrial System: Measurement, Data Analysis and Theoretical Models, Proc. of the Int. School of Physics E.Fermi, Varenna, 1996, Course CXXXIII, edited by G.Cini Castagnoli and A.Provenzale (IOS Press, Amsterdam), p.1-24. [27] Svensmark, H., and Friis-Christensen, E., 1997, Variation of cosmic ray flux and global cloud coverage- a missing link in solar-climate relationships, J.Atmos.Sol.Terr.Phys. 59, 1225-1232. [28] Svensmark, H., 1998, Influence of cosmic rays on Earthís climate, Phys.Rev.Lett., 81, 5027-5030. [29] Mangia, C., De Sanctis, L.V., Ruggiero, L., Zito, G., Zuanni, F., 1991, Un secolo di precipitazioni piovose a Lecce, 4th Workshop Progetto strategico: Clima, ambiente e territorio nel mezzogiorno, ISIATA-CNR, Lecce, 153-161. [30] Buffoni, L., Maugeri, M., Nanni, T., 1999, Precipitation in Italy from 1833 to 1996, Teor. Appl. Climatol. 63, 33-40. [31] G.Russo, C.Eva, C.Palau, A.Caneva and A.Sacchini, 2000, Nuovo Cimento, C23, 39-52. [32] Vautard, R., Yiou, P., and Ghil, M., 1992, Singular Spectrum Analysis: a toolkit for short, noisy, chaotic time series, Physica D58, 95-126. APPENDIX 1 The SSA approach involves 3 basic steps: 1) embedding the time series in a vector space of dimension M (for the choice of M, see Ref.[32]); 2) computing the MxM lag- covariance matrix CD of the data (see the two different approach of Broomhead and King [20] and of Vautard and Ghil, [21]); 3) diagonalising CD : T LD = EDCDED, where LD = diag(l1,l2,...,lM ), with l1 ≥ l2 ≥...≥ lM ≥ 0 and ED is the MxM matrix having the corresponding eigenvectors Ek , k=1,...,M as its columns. For each Ek we construct the time series (of length N-M+1), called k-th principal component (PC), representing the projection of the original time series in the direction determined by the eigenvector Ek (also called empirical orthogonal function, EOF). Each eigenvalue l k gives the variance of the corresponding PC; its square root is called singular value (SV). Having chosen a subset of eigenvalues, it is possible to extract time series of length N, combining PCs; these time series are called reconstructed components (RCs) and they can be superimposed to the original signal. COSMIC RAY MEASUREMENTS IN THE ATMOSPHERE Y.I. Stozhkov, N.S. Svirzhevsky, and V.S. Makhmutov Lebedev Physical Institute, Russian Academy of Sciences, Moscow, Russia Abstract We present the main characteristics of cosmic ray population in the atmosphere and its variability (11-and 22-year solar cycle variations, solar protons originating from powerful solar flares, energetic electron precipitation during geomagnetic disturbances and Forbush decreases of cosmic rays). The experimental data were obtained from the long-term cosmic ray monitoring in the atmosphere from 1957 to now. The relationship between cosmic ray fluxes, and atmospheric processes are also discussed. 1. METHOD OF REGULAR MEASUREMENTS OF COSMIC RAYS IN THE ATMOSPHERE Regular cosmic ray measurements of cosmic rays are carried out with different instruments: ionization chambers, meson telescopes, and neutron monitors on the ground. The idea of cosmic ray monitoring in the atmosphere with radiosondes was suggested by Prof. S.N. Vernov in the middle of fifties and was realized by him and Prof. A.N. Charakhchyan in 1957. Now the cosmic ray monitoring cover a wide range of cosmic ray energy spectrum. It is schematically shown in Fig. 1. 5 SCR 4 GCR 3 atm. cosmic ray flux 2 nm 1 7891011 12 lg E, eV Fig. 1 Schematic view of galactic and solar cosmic ray spectra (GCR, SCR, thick and thin curves accordingly). The dotted vertical lines show the minimal energy of primary particles, which are detected by radiosondes in the atmosphere (E>0.1 GeV, upper arrow labeled atm) and by neutron monitors on the ground level (E>1.5 GeV, arrow with nm). The ground-based ionization chambers and meson telescopes record the primary particles with E>9 GeV. The cosmic ray measurements in the atmosphere are made with standard radiosondes in which the charged particle detectors are Geiger counter (hereafter counter) and telescope consisting of two counters and with 7 mm Al plate between them. Single counter records charged particles (electrons with energy E>0.2 MeV, protons with E>5 MeV) and g-rays with E>0.02 MeV (efficiency <1 %). Telescope records electrons with E>5 MeV, protons with E>30 MeV and is not sensitive to g-rays. For the isotropic angular distribution of particles in the upper hemisphere the geometrical factors of these detectors are 15.1 cm2 and 17.8 cm2 sr. The long-term cosmic ray measurements in the atmosphere have been started at the several latitudes with the different geomagnetic cutoff rigidities Rc in the middle of the last century and they are continued till now [1]. Every day or several times per week balloon flights have been made. Also several sea expeditions had been organized where the measurements of cosmic ray fluxes in the atmosphere in a wide range of Rc had been made. Till now more than 70.000 balloon flights have been performed. In Table 1 the sites and periods of observations are given. The cosmic ray fluxes are measured from the ground level up to 30-35 km. Table 1. The sites and periods of cosmic ray measurements in the atmosphere. Site of observations Geographic Rc, GV Period of observations coordinates Mirny, Antarctica 66∞33¢ S; 93∞00¢ 0.04 03.1963 - present time Tixie 71∞33¢ N; 128∞54¢ 0.4 02.1978 Ð 09.1987 Murmansk region 68∞59¢ N; 33∞05¢ 0.6 07.1957 - present time Norilsk 69∞00¢ N; 88∞00¢ 0.6 11.1974 Ð 06.1982 Moscow region 55∞28¢ N; 37∞19¢ 2.4 07.1957 - present time Alma-Ata 43∞12¢ N; 76∞56¢ 6.7 03.1962 Ð 02.1992 Erevan 40∞10¢ N; 44∞30¢ 7.6 01.1976 Ð 06.1989 Sea expeditions 60∞ N - 60∞ S 0.1-17 1963-65; 1968-72; 1975-76; 1986-87 In the atmosphere the main part of charged particles is secondary ones except of altitudes h≥20 km in polar regions where there are low energy primary protons. Below h<20 km cosmic rays mainly consist of secondary electrons and muons. 2. GALACTIC COSMIC RAYS To study galactic cosmic ray flux variations in different energy intervals the magnetic field of the Earth and the atmosphere are used as separators of charged particles according to their rigidity and energy. As an example in Fig. 2 the data obtained during the flights of radiosondes at Murmansk, Mirny, and Moscow on September 1997 are presented. During 1997 solar activity level was low and cosmic ray fluxes in the atmosphere were maximal ones. In Fig. 3 and 4 the samples of data obtained at the northern and southern latitudes with the different values of Rc during the Antarctic sea expedition of 1986-87 are shown [2]. The several radiosonde launchings were made at the each latitude and averaged data are presented in these figures. One can see a noticeable dependence of cosmic ray fluxes on Rc. Also the atmospheric depth (or pressure) X where maximum fluxes of charged particles, Nm, are observed increases with the growth of Rc. The examples of the time dependencies of charged particle fluxes (monthly averaged values) measured at the polar (northern and southern) and middle latitudes in the stratosphere and troposphere are given in Fig. 5 and 6. The period of observations covers ~19-23 solar activity cycles. 40 35 30 25 20 h, km 15 10 5 0 0 1000 2000 3000 4000 N, min-1 Fig. 2. The count rate of single counter vs. altitude in the atmosphere: at the northern polar latitude, Murmansk region, Rc=0.6 GV (the radiosonde flights on 2 and 4 September 1997 - open circles and black points, accordingly); at Mirny in the Antarctic, Rc=0.04 GV (the flights on 3 and 8 September - open and black triangles, accordingly); at the middle latitude, Moscow region, Rc=2.4 GV (the flight on 3 September - open squares). 3.5 3.0 0.6 north 2.5 -1 2.4 s 2.0 -2 1.5 5.3 N, cm 1.0 6.7 0.5 10.7 13.7 0.0 110100 1000 X, g cm-2 Fig. 3. The cosmic ray fluxes vs. atmospheric pressure X measured at different latitudes of the northern hemisphere during the sea expedition of 1986-1987. The values of Rc in GV are shown for each curve. The vertical bars show standard errors. 3.5 3.0 0.04 south 2.5 1.7 -1 s 2.0 2.4 -2 1.5 3.4 N, cm 5.4 1.0 7.3 0.5 10 13.7 0.0 110100 1000 X, g cm-2 Fig. 4. The cosmic ray fluxes vs. atmospheric pressure X measured at different latitudes of the southern hemisphere during the sea expedition of 1986-1987. The values of Rc in GV are shown for each curve. The vertical bars show standard errors. From Fig. 6 it is seen that the cosmic ray latitude effect between polar and middle latitudes disappears in the troposphere that is the cosmic ray fluxes observed at these latitudes are equal. 3.5 3.0 -1 2.5 s -2 2.0 N, cm 1.5 1.0 55 65 75 85 95 Year (after 1900) Fig. 5. Time dependence of monthly averaged cosmic ray fluxes in the stratosphere at h=31 km (X=10 g/cm2) measured at the northern and southern polar latitudes (Rc=0.6 and 0.04 GV, upper solid and dotted curves, accordingly) and at the middle latitude (bottom gray curve, Rc=2.4 GV). 1.3 -1 1.1 s -2 N, cm 0.9 0.7 55 65 75 85 95 Year (after 1900) Fig. 6. Time dependence of cosmic ray fluxes averaged per month in the troposphere at h=10.5 km (X=250 g/cm2) measured at the northern polar latitudes (Rc=0.6 GV, solid curve) and at the middle latitude (dotted curve, Rc=2.4 GV). 2 At each atmospheric pressure level, X, g/cm , only particles with the energy E>Ea (or rigidity R>Ra) where Ea is the atmospheric cutoff energy can contribute to the count rate of our detectors. The atmospheric cutoff Ea or Ra is defined by the characteristics of nuclear interactions of primary cosmic rays with air atoms. From the latitude measurements (at the different Rc) one can get the values of atmospheric cutoff as a function of X. In Fig. 7 the relationship of Ra. and X is presented. 10 8 6 , GV a R 4 2 0 10 100 1000 X, g/cm2 Fig. 7. The atmospheric cutoff Ra. vs. atmospheric pressure X. Open points were obtained from the sea expedition data and black points Ð from the long-term data obtained at the stationary sites (see Table 1 and Fig, 3, 4). Solid curve is the approximation: 0.8 2 Ra.=0.04X where R is in GV and X is in g/cm . Thus, the measurements of cosmic ray fluxes at the different atmospheric depths give the information on the variations of primary cosmic ray integral fluxes from R>0.5 GV at Xª10 g/cm2 up to R>(9-10) GV at sea level. R =0.04 and 0.6 G 3.5 c 3.0 -1 s -2 , cm 2.5 max N Rc=2.4 GV 2.0 _ _ + + + 1.5 55 60 65 70 75 80 85 90 95 100 Year (after 190 Fig.8. Monthly values of Nmax (maximum cosmic ray fluxes in the atmosphere) recorded at the Murmansk region, Rc=0.6 GV (upper solid curve), at Mirny station, Antarctic, Rc=0.04 GV (upper gray curve) and at Moscow region, Rc=2.4 GV (bottom curve). The narrow vertical stripes between two dotted lines show the periods of solar polar magnetic field inversions and +/- signs denote the magnetic field polarity of the solar northern polar region. In Fig. 8 the long-term experimental data on maximum fluxes of cosmic rays in the atmosphere obtained at the latitudes with Rc=0.04, 0.6, and 2.4 GV are depicted. In this figure and Figs. 5 and 6 the 11-year changes of cosmic ray fluxes are seen: in 1965, 1977, 1987, and 1997 the Nmax values are maximum ones and in 1957, 1970, 1982, 1991 they are minimal. The peaks observed at the northern polar and middle latitudes in 1962-63 were due to the radioactivity from the nuclear explosions in the atmosphere. In Fig. 8 (also in Figs. 5 and 6) the 22-year solar magnetic cycle is seen in the time dependence of Nmax: during the negative phases of solar magnetic cycles (~1960-70 and ~1980-90) cosmic ray time dependence shows a peaked form and it has a plateau during the positive phases (~1970-80 and ~1990- 2000). The difference in cosmic ray drift current directions in the heliosphere during positive and negative phases of the solar magnetic cycle explains the peaked and smoothed curves observed [3]. The amplitude of the 11-year solar cycle variations decreases with the growth of atmospheric pressure X. In Fig. 9 the changes of yearly averaged cosmic ray minimum fluxes, Nmin, observed in the 11- year solar cycles relative to 1965 are shown as a function of X: A(x)=[100(N65(x) - Nmin (x)]/N65(x), % . The N(x) value had a maximum in 1965 (see Fig. 8). The value A decreases with the increase of X and at X>600 g/cm2 becomes about 3 %. In June- August, 1991 the absolute minimum cosmic ray fluxes were recorded for the whole period of the observation from 1957 till present time (see squares in Fig.9).. 60 (7-9).59 9 (7-9).70 8 50 (10-12).82 (6-8).91 7 40 6 , 5 a 30 R 4 A, % 20 3 2 10 1 0 0 0 200 400 600 800 X, g/cm2 Fig. 9. The amplitude of 11-year cosmic ray changes relative to 1965 vs. atmospheric pressure X. The periods considered (months and year) correspond to the minimum cosmic ray fluxes and are given in the insert of this Figure. Cosmic ray fluxes were averaged for these periods. The atmospheric cutoff Ra is shown by solid line. -0.1 3.5 0 0.1 0.2 -1 3 s 0.3 -2 0.4 , cm 2.5 0.5 max solar activity N 0.6 2 0.7 0.8 0.9 1.5 55 65 75 85 95 Year (after 190 Fig. 10. Time dependence of maximum cosmic ray flux in the polar atmosphere, Nmax (black curve) and solar activity level (gray curve) defined as h/j where h is sunspot number and j is sunspot average helio-altitude. The monthly averaged data smoothed with the period of T=3 months were used. The values of cosmic ray fluxes N(x) in the heliosphere and, in turn, in the atmosphere are defined by solar activity level. The close relationship is observed between N(x) and solar activity parameter (h/j) where h is a sunspot group number, j is sunspot averaged helio-altitude [4]. This relationship is demonstrated in Fig. 10. The correlation coefficient is R(Nmax, h/j) = Ð0.83±0.03. The galactic cosmic ray modulation in the heliosphere is produced by magnetic irregularities of interplanetary magnetic field (IMF). In turn, the value of IMF and its irregularity density are defined by solar activity. The density of these irregularities increases with the growth of IMF strength. So, we can expect a relationship between IMF strength and cosmic ray fluxes in the heliosphere as well as in the atmosphere. This relationship is given in Fig. 11. The correlation coefficient R(Nmax, IMF)= -0.71±0.04. 4 3.5 5 -1 3 6 s -2 7 IMF, , cm 2.5 8 max N 9 2 10 1.5 11 55 65 75 85 95 Year (after 190 Fig.11. Time dependence of maximum cosmic ray flux in the polar atmosphere, Nmax (black curve) and IMF (gray curve). The monthly averaged data smoothed with the period of T=3 months were used. The data on IMF were taken from INTERNET: http://nssdc.gsfc.nasa.gov/omniweb/. 30 25 20 15 Start time, UT, date h, km GCR 10 0822 11.09.0 1053 11.09.0 5 1942 11.09.0 0326 11.10.0 0 0.1 1.0 10.0 100.0 N, cm-2 s-1 Fig. 12. Cosmic ray fluxes recorded in the atmosphere at Mirny in the Antarctic during the solar proton event on 9 November 2000. Solid curve is a charged particle flux background produced by galactic cosmic rays in the atmosphere (GCR). Different symbols show the data obtained during the different flights of radiosondes (the date and start time of launchings are given in the insert).. 3. SOLAR COSMIC RAYS, PRECIPITATION, AND RADIOACTIVITY Since the beginning of cosmic ray measurements in the atmosphere in 1957 several tens of solar proton events were recorded (e.g. [5]). As a rule such events are observed in the polar atmosphere where Rc are rather low (see Table 1). During solar proton events total cosmic ray flux at high altitudes in the atmosphere increases in several (sometimes in tens) times. As example, the solar proton fluxes generated by solar flare on November 9, 2000 and recorded in the Earth’s atmosphere are given in Fig. 12. Solar proton fluxes were observed in the atmosphere at h>17 km and their values increase with altitude. 100 1 3 -1 sr 10 -1 2 s -2 1 J(>E), cm 4 0.1 100 1000 E, MeV Fig. 13. The energy spectra of solar protons on 9 November 2000. The start times of radiosonde launchings and exponents g of solar proton energy spectra I(>E)~E-g are given below: 1 ÐDate - 11.09.00, Start time (UT) - 8:22, g=7.2; 2 Ð Date - 11.09.00, Start time (UT) - 10:53, g=5.8; 3 Ð Date - 11.09.00, Start time (UT) - 13:36, g=7.8; 4 Ð Date Ð 11.10.00, Start time (UT) - 19:42, g=4.9. From these data the fluxes and energy spectra of solar flare particles in the energy range of E=100- 500 MeV were obtained. They are depicted in Fig. 13. In this solar proton event the particles with E>500 MeV were not observed and this event was not recorded by neutron monitors. We note that the observed solar proton “soft” energy spectra could be due to the additional acceleration of particles in the interplanetary space by shock waves as it was happened in the past during the solar proton events in August 1972 [6]. 200 10 8 150 6 100 4 sunspot number 50 2 solar proton event number 0 0 1955 1965 1975 1985 1995 Year Fig. 14. The time dependences of yearly number of the solar proton events recorded in the atmosphere (black points and right axes) and yearly sunspot number (open points and left axes). For the 45-year period of cosmic ray monitoring in the atmosphere 105 solar proton events were recorded. In Fig. 14 the time dependence of the yearly number of these events and solar activity level (sunspot number) are shown. It is seen that the solar proton events mainly occur during the ascending and descending phases of solar activity. 35 35 30 30 25 25 20 20 h, km 15 h, km 15 10 10 Start time UT Date Start time UT Date 5 1250 05.03.0 5 1250 05.03.0 0930 05.05.0 0930 05.05.0 0 0 0 1234 0123 N, cm-2 s-1 N, cm-2 s-1 Fig. 15. Precipitation of high energy electrons into the northern polar atmosphere recorded by single counter on 5 March 2000 (left panel, black points). The background from galactic cosmic rays is shown by open points. The telescope recorded galactic cosmic ray background only (right panel). The inserts show the dates of radiosonde flights and launching times. During the geomagnetic disturbed periods in the polar atmosphere at high altitudes high-energy electron precipitation events are detected [7, 8, 9]. Our northern polar station in Murmansk region is near the polar oval (McIlwain’s parameter L=5.6) where the precipitations are observed rather often. In Fig. 15 the example of precipitation detected on November 2000 is shown. The counter records secondary g-rays produced by precipitating electrons in the atmosphere.. But at the same time the telescope records only the background from galactic cosmic rays and it allows to separate precipitation from solar proton events. It is significant that g-rays recorded in the atmosphere at h=25-35 km are produced by the precipitating electrons with E>several MeV. The time dependence of the yearly precipitation number and the sunspot number are given in Fig. 16. Into the precipitation data the corrections for patrol efficiency were introduced. The data obtained show that the precipitation take place most often during the descending phase of solar activity (remind that solar proton events are observed most often during the ascending and descending solar activity phases, see Fig. 14). This fact was established earlier in other papers [10]. The total number of precipitation recorded at the station in Murmansk region during 1957-2000 equals 549 events. For almost the same period (1963-2000) at the Antarctic station Mirny (66∞33¢ S; 93∞00¢; Rc=0.04 GV; Lª11) 10 precipitation events were recorded only. In Table 2 the yearly precipitation numbers for 20 - 23rd solar cycles are given. The last line in this Table includes the ascending phase of 23rd solar activity cycle only. The regular monitoring of charged particle fluxes in the atmosphere provides the prompt control of radiation conditions and allows to detect radiation clouds from nuclear explosions or nuclear plant failures. In Fig. 17 and 18 the observations of radioactive clouds in the polar northern atmosphere and over Moscow are shown as an examples. The excesses of particles over the cosmic ray background recorded by single counter were due to the radioactivity particles. The telescope data showed normal count rate from cosmic rays only. A powerful surface nuclear explosion was produced near lake Lobnor in China on the 14th of October 1970. In the atmosphere radioactive clouds were observed near Murmansk on October 25-26 (see Fig. 17) and on November 11-12, near Moscow and Alma-Ata on November 5 and 6. In the polar region it was observed in the altitude range hª15-25 km and charged particle flux was increased in 7-8 times in comparison with the cosmic ray background. At the first registration the cloud had the shape of a disk with vertical size of nearly 4 km and horizontal size along the wind direction of ~1100 km. The maximal activity was equal ~10-3 Bq/cm3 as measured at hª18-22 km. After 26 October the radioactive cloud passed away from the observation site. A radioactive cloud in the atmosphere over Moscow was recorded on 12-14 April 1993 (see Fig. 18). The cloud was seen at hª10-30 km and had the horizontal extension of nearly 1000 km and maximum activity ~10-4 Bq/cm3. We do not know this cloud origin but on 6 April the failure at the chemical plant in Siberia (Seversk town near Tomsk) occurred [11]. A more dense cloud was observed over Moscow on the 27th of October 1999, the maximum activity being 7◊10-3 Bq/cm3 at h>15 km. The cloud horizontal extension was about 200 km. The source of this last cloud is unknown also. 100 200 80 150 60 100 40 50 sunspot number 20 precipitation number 0 0 1955 1965 1975 1985 1995 Year Fig. 16. The time variations of the yearly number of precipitation (black points) and sunspot number (open points). The observations were made at the northern polar station in Murmansk region (68∞59¢ N; 33∞05¢; Rc=0.6 GV; L=5.6) during the time interval of (6-12) UT. Table 2. The precipitation number in different solar activity cycles. solar cycle number, sunspot number precipitation Precipitation period per year number (total) number per year 20 (1964-1975) 58.8 144 12 21 (1976-1985) 82.9 140 14 22 (1986-1995) 78.5 118 12 23 (1996-2000) 46.9 85 17 30 25 20 15 h, km background 10 1602 LT 25 Oct. 197 1802 LT 25 Oct. 197 5 2207 LT 25 Oct. 197 1157 LT 26 Oct. 197 0 051015 20 N, cm-2 s-1 Fig. 17. Charged particle fluxes in the northern polar atmosphere detected by single counter. At h>15 km the excess of flux over the cosmic ray background (solid gray line) is due to radioactive cloud particles produced by the nuclear explosion in China on 14 October 1970. In the insert the legend on the balloon start time is given. 40 30 20 h, km 10 0 012345 N, cm-2 s-1 Fig. 18. The count rate of single counter vs. altitude in the atmosphere over Moscow on 12-14 April 1993:æ background from galactic cosmic rays; charge particle flux measurements on 13 April, launching local time 0830 LT (◊), on 13 April, 1430 LT (D); on 14 April, 0830 LT ( );. 4. COSMIC RAY FLUXES AND ATMOSPHERIC PROCESSES 10 If one compares the flux of solar electromagnetic radiation falling on the top of the atmosphere (Fsunª10 -2 -1 2 -2 -1 erg m s ) with the flux of cosmic ray energy (FCRª10 erg m s for particles with energy E≥0.1 GeV) the evident conclusion could be made: the influence of charged cosmic ray particles on the processes in the atmosphere is negligible in comparison with influence of the electromagnetic radiation coming from the Sun. However, let us imagine for a moment that cosmic rays stopped to intrude into the Earth’s atmosphere. The ion production will be aborted and the global electric circuit will be destroyed. The production of thundercloud electricity and lightning will be over. The cloud area will be decreased and precipitation level will fall down. The cosmic rays with energy E=0.1-15 GeV carry about 60 % of all cosmic ray energy and these particles constitute about 95 % of all cosmic ray flux. These particles undergo the influence of the geomagnetic field in such way that the fluxes of primary cosmic rays at polar latitudes is higher than the ones at equatorial regions as much as ~30-35 times. In the atmosphere this difference is about 4 times. Below some aspects of influence of charged particle fluxes on atmospheric processes are considered (see also [12]). In our analysis we use the experimental data of the long-term measurements of cosmic ray fluxes at the different atmospheric depths (from the Earth’s surface up to 30-35 km) and at the different latitudes. 4.1 The global electric circuit and ion production It is well known that the Earth has a negative electric charge about 6¥105 C and the strength of electric field produced by this charge near the Earth’s surface measured during fair-well weather equals Eª-130 V/m (directed to the Earth’s surface). The value of average current flowing between equalizing layer to be in the ionosphere at the altitude hª55-80 km and the Earth’s surface is Iª10-12A/m2 [13, 14]. The light ions provide this current in the atmosphere. The ions are produced by cosmic particles (radioactivity of soil also gives ions but only in the lower atmosphere at h< 3 km). If cosmic ray flux changes the ion density the air conductivity changes also. The lightning in thunderstorms and precipitation form another branch of the closed electric circuit charging the Earth by negative electricity and providing electric current from the Earth to the ionosphere. The sketch of global electric circuit is given in Fig. 19 (see, e.g. [16]). h =60 km Ip Rp Im, Rm Earth Ig Fig. 19. The sketch of the global electric circuit: h=60 km Ð equalizing layer; Ip, Rp and Im, Rm Ð atmospheric electric currents and resistances in the atmosphere at polar and middle latitudes, accordingly; Ig Ð current of thunderstorms and precipitation charging the Earth by negative electricity. The equation describing the relation between ion production rate, q, and their recombination in the atmosphere under quasi-state conditions is usually taken in the form q(h)=a(h) [n(h)]2, (1) where n is ion concentration, a is recombination coefficient, h is atmospheric altitude [17]. Using the experimental data on cosmic ray fluxes and ion concentrations in the atmosphere one can test the validity of this equation. In Fig. 20 the ion concentrations, n, and the charged particle fluxes, N, measured at several latitudes vs. altitude are presented [18]. From the experimental data on ion concentration n and cosmic ray flux N one can get that the ion production rate q is proportional to charged particle flux: q(h)=msN(h), where m and s are the number of air particles per cm3 and ionization cross-section. The values of m and s are the same for different latitudes and depends on the altitude only. It isn’t true for the case of polar latitudes and h>20 km where s is increased. At h<20 km the value of s equals 2*1018 cm2 within 10-15 % for all latitudes. 17.3 5.3 3 17.3 5.6 3.4 0.04 30 0.6 30 20 20 h, km h, km 10 10 0 0 0123 012 n, 103 cm-3 N, cm-2 s-1 Fig. 20. Ion concentration n (left panel) and cosmic ray flux N (right panel) as a functions of altitude h in the atmosphere at the latitudes with the geomagnetic cutoff rigidities Rc=17.3, 5.6, 5.3, 3.0, 3.4, 0.6 and 0.04 GV. Horizontal bars show the standard errors. Let us consider the measurements of n and N performed at two different latitudes. According to the expression (1), we can construct the following ratio: 2 2 [q1(h)/q2(h)]=[a(h) n (h)]1/[a(h) n (h)]2, (2) where the subscripts 1 and 2 correspond to the latitudes with different geomagnetic cutoff rigidities Rc1 and Rc2. Substituting msN(h) instead of q and taking (ms)1= (ms)2 and a1ªa2 (these suggestions are fulfilled in the atmosphere rather well) one can get 2 [N1(h)/N2(h)]=[n1(h)/n2(h)] . (3) 0.8 1 0.6 0.4 2 Ratio 0.2 3 0.0 51015202530 h, km Fig. 21. The ratio of cosmic ray fluxes (curve 1), ion concentrations (curve 2) and squared ion concentrations (curve 3) as a function of altitude. These values were calculated from the experimental data obtained at the equatorial (Rc=17.3 GV) and middle (Rc=3.3 GV) latitudes (see Fig.20) without any normalization of the data. The standard errors are given by vertical bars. In Fig. 21 the ratios of charged particle fluxes (curve 1-open points), ion concentrations (curve 2- dark points), and squared ion concentrations (curve 3-crosses) obtained from the experimental data presented in Fig. 20 are given. The data obtained in the equatorial (Rc=17.3 GV) and middle latitudes (Rc=3.3 GV) were used. It is seen that the ratio of cosmic ray fluxes (curve 1) coincides with ion concentration one (curve 2) and differs significantly from squared ion concentration ratio (curve 3). The details of such consideration are given in [18]. Thus, from this analysis the important conclusion must be made that the ion balance in the atmosphere under quiet conditions is described by linear equation (not quadratic one) q(h)=b(h) n(h), (4) where b(h) is the linear recombination coefficient. From the available experimental data on cosmic rays in the atmosphere and light ion concentrations the value of b(h) and q(h) can be calculated for any site of the Earth and any level of solar activity. The ion production rate q can be written as q(h)=N(h) s(h) r(h)/M, (5) where N(h) is cosmic ray flux at the altitude h, s is the ionization cross-section in air, r(h) is the air density and M is the average mass of air atom. The relationship between atmospheric electric current J, electric field strength E and conductivity l is J=l(h) E(h)=n(h) k(h) E(h), (6) where k(h) is the mobility of light ions at the altitude h. Thus, using the expressions 4, 5 and 6 one can find J=N(h) s(h) r(h) k(h) E(h)/[M b(h)]. (7) 2 0.66 N -2 1.6 -1 0.62 s A m -2 -12 J N, cm J, 10 1.2 0.58 0.8 0.54 1965 1970 1975 1980 1985 Year Fig. 22. The yearly average values of atmospheric electric current J(h) (from [19]) and cosmic ray flux N(h) at hª8 km in the polar region. On the right side of this equation all values are constant except cosmic ray flux N(h) and electric field strength E(h). If one supposes that E(h) is constant or weakly changes in the periods of fair-well weather then there is the linear relationship of cosmic ray flux N(h) and atmospheric electric current J(h). Such conclusion is confirmed by the experimental data showed in Fig.22. The data on J(h) were taken from [19]. The correlation coefficient between J(h) and N(h) is positive and equals r(J, N)= +0.77±0.10. The correlation of atmospheric electric current and solar activity level (sunspot number W) is low, r(J,W)=-0.32±0.22. 4.2 Thundercloud electricity and lightning production In 1920 Wilson put forward fascinating idea suggested that thunderstorms act as a global generator of electric current maintaining the Earth’s electric charge [20]. Since the experimental evidences supporting this hypothesis were obtained (see references in [14]). However, the mechanisms of thunderstorm electricity production (separation of negative and positive charges in thundercloud) and lightning generation are not clear till present time, although there are a number of hypotheses on the thundercloud electricity origin (see, e.g., [21-23]). Cosmic rays could be responsible for the thunderstorm electrification [24]. Secondary charged particle fluxes generated in the atmosphere by primary cosmic rays are the only source of positive and negative ion production in the atmosphere at h>(2-3) km. The problem consists in the spatial separation of negative and positive ions in the process of thundercloud formation. The thunderclouds are formed from ascending wet air mass when the fronts of cold and warm air meet each other. The air masses contain heavy ions (charged aerosols) because light ions produced by cosmic rays adhere to neutral heavy particles. As it is known from the observations the concentration of aerosols has a maximum in the low atmosphere near the Earth’s surface and its value is ~2¥104 cm-3. The half of these particles carries out the positive or negative electric charges [25]. Ascending air mass picks up the aerosols. During ascending air mass is cooled and processes of condensation of water molecules on neutral and charged aerosols take place. The condensation rate depends essentially on the charge presence and its sign. Namely, negative charged aerosols grow faster than positive ones as much as ~104 times [26, 27]. The rapid growth of aerosols with negative charge makes them heavy and their lift with the rising air mass is stopped at the low altitudes. At the same time aerosols with positive charge continue to rise with ascending wet air mass and stop their rising at higher altitudes than negative charged aerosols. In this way the spatial separation of electric charges inside the cloud occurs (in detail see [24]). Inside the thundercloud the strength of electric field can grow up to Eª3 kV/cm and the distance between separated positive and negative charges is roughly estimated as Dhª3-4 km. The high value of E is observed under thundercloud also. But the observed values of E are much less than the puncture voltage at the altitudes where thunderclouds exist (hª2-7 km). At hª3 km the value of puncture voltage is 15-30 kV/cm [28]. In [29] Ermakov put forward the idea that in such electric fields the discharges (lightning) are produced by extensive air showers arising from high energy cosmic ray particles with e =1014-10 15 eV. These high-energy cosmic rays interact with nuclei of ambient air and give rise to many thousands of charged secondaries. Along ionized tracks of these secondary particles in a strong electric field the avalanches develop and propagate. The high energy cosmic ray particle flux is enough to explain the number of lightning observed. As cosmic rays hit the Earth’s atmosphere accidentally in all directions the lightning arise by chance also. There is another mechanism of lightning production suggested by Gurevich [30, 31] in which relativistic electron is accelerated in the electric field of thundercloud and produces avalanche. In the process of thundercloud formation one can recognize initial, maturity and decay phases. In Fig. 23 these phases and the processes of thundercloud electricity and lightning production are shown schematically [24]. _ J ______9 + + + + + + + + + + + + + + + + + + _ + + + + + _ _+ _+ _+ _ + _ + + + + + _ _ _ _ _ + + 5 + + + _6 ______5_ _ _ + + + + + + 1 4 1 ______2 + + + + + + + _ _ _ _ _ 2 + + + + + + 3 + + + + + + 3 7 + + + + _ + _ + + _+ _ _+ 8 + a + _ + + 10 b c Fig. 23. The phases of thundercloud formation: a Ð initial phase; b Ð maturity phase; c Ð decay phase. Notations: 1 and 2 Ð warm and cold fronts of air; 3-ascending flux of wet air with ions; 4 and 5 Ð extensive air showers produced by primaries with energies e ≥1014 eV and e ≥1015 eV accordingly; 6 Ð intracloud lightning; 7 Ð cloud-to-ground lightning; 8 Ð ground-to-cloud lightning; 9 Ð the negative screen layer; 10 Ð positive charge on the cloud base; J Ð current of negative ions from the ionosphere to the top of the thundercloud. 4.3 The relationship between cosmic rays and other atmospheric phenomena There are several publications in which the changes of cosmic ray fluxes are considered to be responsible for some processes in the atmosphere. Some of such phenomena are listed by Tinsley [16]. Below we analyze the relationships of cosmic ray fluxes with cloudiness, precipitation and lightning. The influence of charged particle fluxes on cloudiness was found by Veretenenko and Pudovkin [32]. They found that the value of cloudiness reduced when cosmic ray fluxes in the interplanetary space and in the atmosphere decreased during so-called Forbush-effects of cosmic rays. As was shown by Stozhkov et al. [33, 34] during Forbush-effects the value of precipitation decreased also. In contrast, when the ionization level is increased due to the invasion of solar flare protons into the atmosphere the precipitation level also increased. These results were obtained from the analyses of the precipitation data recorded at the numerous meteorological stations located in Brazil and the Former Soviet Union. More than two hundreds of Forbush-effects and several tens of solar flare events were analyzed. In the analyses the superposed epoch method was applied. Figures 24 and 25 demonstrate the changes of precipitation in the cases of decreases and increases of cosmic ray fluxes in the atmosphere [33, 34]. The value of precipitation level decrease obtained by superposed-epoch method for more than 70 events of Forbush effects is D0= -(17.4±2.7) %. The probability of the occasional appearance of effect is less than 10-4 if the values of D have a normal distribution. The results on relative increase of rainfall level of D during solar proton events were obtained by superposed-epoch method for more than 53 events of solar proton enhancements. The amplitude of positive increase is D0=(13.3±5.3) %. The probability of effect appearance by chance is less than 10-2. D, % 10 5 0 -35 -25 -15 -5 5 15 25 35 -5 Days -10 -15 -20 Fig. 24. The changes of the daily precipitation level, D, %, relative to mean value evaluated from precipitation data during 1 month before (-30 to –1 days) and 1 month after (1 to 30 days) Forbush decrease event. The day “0” correspond to the Forbush decrease main phase. The precipitation data used in the analyses were obtained in the Former Soviet Union and Brazil. D, % 15 10 5 0 -30 -20 -10 0 10 20 30 -5 Day -10 -15 Fig. 25. The changes of the daily precipitation level, D, %, relative to mean value evaluated from precipitation data during 1 month before and 1 month after solar proton events recorded by ground-based neutron monitors (“0”-day). The link of cosmic ray intensity and global cloud coverage was found by Svensmark and Friis- Christensen [35]. Their results demonstrate the relationship between charged particle fluxes on the Earth’s surface and cloudiness during long-term cosmic ray modulation in the 11-year solar activity cycle. When cosmic ray flux increases cloudiness increases and one can expect that the number of thundercloud (or thundercloud coverage) will increase also. The ion production rate and ion concentration in air grow; the total electric charge in the thunderclouds increases. Thus, the number of lightning has to grow and the relationship between cosmic ray flux or ion production rate and thundercloud discharge number has to take place. Now there are the long-term experimental data on lightning flashes over the United States [36] and the link of lightning number and cosmic ray fluxes can be checked. In Fig. 26 the relationship of lightning number L with ion production rate q is shown. The correlation coefficient between these values is r(L, q)=+0.85±0.09. The values of q were calculated from the data on cosmic ray flux measured in the atmosphere at the middle latitudes. 12 30 s) 2 11 25 q 10 events/year 20 6 ion pairs/(cm 15 9 6 L, 10 L q, 10 10 8 1988 1992 1996 2000 Year Fig. 26. The yearly number of lightning L detected in United States in 1989-1998 (black points, from [36]) and ion production rate q in the air column (h=2-10 km) of the middle latitudes (open points). 4.4 Artificial influence on precipitation The results obtained by Veretenenko and Pudovkin [32], Stozhkov et al. [33], Svensmark and Friis- Christensen [35], Ermakov and Stozhkov [24] show clearly the important role of charged particle fluxes on the cloud, thundercloud formation, and precipitation processes. In the lower atmosphere the changes of cosmic ray flux during Forbush effects (decrease of cosmic ray flux) or solar proton events (increase of flux) is about (2-15) %. In the first case the decrease of precipitation level is observed, in the second one the growth of precipitation takes place (see Figs. 24 and 25). It is possible to increase the flux of charged particles in the lower atmosphere using an electron accelerator onboard airplane. The accelerated electrons can irradiate the cloud increasing ionization level inside the cloud. In turn, it could increase precipitation. The modern linear machines accelerating electrons up to the energy of several tens of MeV have a suitable weight and the energy consumption to be installed onboard airplane, e.g. a machine accelerating electrons up to energy 10 MeV with the current 10 mA has the weight about 1 ton, sizes of 3¥0.5¥1 m3 and the energy consumption of ~1 kW. Let us consider a cloud of 3¥3¥2 km3 in sizes, the top of which is at the altitude ~3 km. The flux of cosmic ray secondaries (mainly relativistic electrons and muons) falling on the upper surface of such cloud is ~7¥109 particles/s. The total energy released by these particles inside the cloud to ionize air atoms equals ~4.5¥106 erg/s. In contrast, the total energy released by the accelerated electrons (~6¥1013 electrons/s) is ~109 erg/s. Thus, the accelerator with the parameters given above could increase the ionization inside the cloud in 10-100 times in comparison with the natural background produced by cosmic rays. In comparison with the methods of artificial influence on the clouds used in practice [37, 38] the irradiation of clouds by accelerated particles is safe because the accelerated electrons and their secondaries will be absorbed in ambient air because of its energy losses. The proposed method could be useful to struggle with droughts and downpours causing floods. 5. CONCLUSION We present the main characteristics of cosmic ray population in the atmosphere and its variability. The experimental data were obtained from the long-term cosmic ray monitoring in the atmosphere from 1957 to now. The main features of cosmic ray variability are the following: ¥ 11-and 22-year solar cycle variations; ¥ solar proton events originating from powerful solar flares; ¥ energetic electron precipitation during geomagnetic disturbances and Forbush decreases of cosmic rays; The cosmic ray monitoring in the atmosphere allows to detect radioactive clouds producing by nuclear explosions or failures at atomic plants. Cosmic ray particle fluxes play important role in many atmospheric processes and only now this role begins to be elucidated. The thundercloud electricity and lightning production, cloud formation, influence on the value of global cloudiness and precipitation on the short (days) and long (11-year solar activity cycle) time scales, operation of global electric circuit and long-scale global climate changes depend on the values of cosmic ray flux. ACKNOWLEDGEMENTS We are very grateful to our colleagues who made hard work to get the experimental data on the long-term cosmic ray variations in the atmosphere. This work is partly supported by Russian Foundation for Basic Research (grants No. 01-02-31005 and No. 99-02-18222). REFERENCES [1] A.N. Charakhchyan, G.A. Bazilevskaya, Y.I. Stozhkov, and T.N. Charakchyan, Cosmic rays in the atmosphere and in space near the Earth during 19 and 20 solar activity cycles, Tr. Fis. Inst. im P.N. Lebedeva, Russian Akad. Nauk, Nauka, 83 (1976) 3 (in Russian). [2] A.E. Golenkov , A.K. Svirzhevskaya, N.S. Svirzhevsky, and Y.I. Stozhkov, Cosmic ray latitude survey in the stratosphere during the 1987 solar minimum, Conf. Pap., Int. Cosmic Ray Conf., XXIst, 7 (1990) 14. [3] H. Moraal, Observations of the eleven-year cosmic-ray modulation cycle, Space Sci. Rev., 19 (1999) 845. [4] Y.I. Stozhkov and T.N. Charakhchyan, On the role of the heliolatitudes of the sunspots in the 11- year galactic cosmic ray modulation, Acta Physics Academial Sciantiarum Hungaricae, (Suppl. 2), 29 (1970). [5] G.A. Bazilevskaya, M.B. Krainev, Yu.I. Stozhkov , A.K. Svirzhevskay, and N.S. Svirzhevsky, Long-term Soviet program for the measurement of ionizing radiation in the atmosphere, Journ. Geomag. and Geoelectr., 43 (1991) 893. [6] G.A. Bazilevskaya., A.N. Charakhchyan, Y.I. Stozhkov, and T.N. Charakhchyan, The energy spectrum and the conditions of propagation in the interplanetary space for solar protons during the cosmic ray events on August 4 to 9, 1972, Conf. Pap., Int. Cosmic Ray Conf., XIIIrd, Denver, USA, 2 (1973). [7] V.S Makhmutov, G.A. Bazilevskaya, A.I. Podgorny, Y.I. Stozhkov, and N.S. Svirzhevsky, The precipitation of electrons into the Earth's atmosphere during 1994, Proc. 24 ICRC, Italy, Rome, 4 (1995) 1114. [8] G.A. Bazilevskaya and V.S. Makhmutov, The electron precipitation into the atmosphere according to cosmic ray experiment in the stratosphere, Izv. AN SSSR, Ser. Fiz., 63 (1999) 1670 (in Russian). [9] V.S. Makhmutov, G.A. Bazilevskaya, M.B. Krainev, Characteristics of energetic electron recipitation into the Earth's polar atmosphere and geomagnetic conditions, Adv. Space. Res., (2001) (in press). [10] G.D. Reeves, Relativistic electrons and magnetic storms: 1992-1995, Geoph. Res. Lett., 25 (1998) 1817. [11] G.A. Bazilevskaya, A.K. Svirzhevskay, N.S. Svirzhevsky, Y.I. Stozhkov, Radioactive cloud in the atmosphere at Moscow site on 12-14 April 1993, Kratkie soobsheniya po fizike, Moscow, Lebedev Instituite, 7-8 (1994) 36 (in Russian). [12] Y.I. Stozhkov, V.I. Ermakov, and P.E. Pokrevsky, Cosmic rays and atmospheric processes, Izv. Russian Akad. Nauk, ser. fiz., 65 (2001) 406 (in Russian). [13] J. Alan Chalmers, Atmospheric Electricity, Pergamon press (1967). [14] R. Reiter, Phenomena in Atmosphere and Environmental Electricity, Amsterdam, Elsvier (1992). [15] R. Markson, Solar modulation of atmospheric electrification and possible implications for the Sun-Weather relationship, Nature, 273 (1978) 103. [16] Brain A. Tinsley, Correlations of atmospheric dynamics with solar wind-induced changes of air- earth current density into cloud tops, Journ. Geophys. Res., 101 (1996) 29,701. [17] L.B. Loeb, Basic Processes of Gaseous electronics, New-York (1960). [18] V.I. Ermakov, G.A. Bazilevskaya, P.E. Pokrevsky, and Y.I. Stozhkov. Ion balance equation in the atmosphere, Journ. Geoph. Res., 102 (1997) 23,413. [19] R.G. Roble, On solar-terrestrial relationships in the atmospheric electricity, Journ. Geoph. Res., 90 (1985) 6000. [20] C.T. Wilson, The maintenance of the Earth's electric charge, Observatory, 45 (1922). [21] E.R. Williams, Electricity of thunderclouds, Scientific American, 1 (1989) 34. [22] M.B. Baker and J.G. Dash, Mechanism of charge transfer between colliding ice particle in thunderstorms, Journ. Geoph. Res., 99 (1994) 10,621. [23] V. Brooks, C.P.R. Saunders, An experimental investigation of the inductive mechanism of thundercloud electrification, Journ. Geoph. Res., 99 (1994) 10,627. [24] V.I. Ermakov and Y.I. Stozhkov, New mechanism of thundercloud electricity and lightning production, Proc. 11-th Intern. Conf. Atmospher. Elect., Alabama, USA (1999) 242. [25] P.N. Tverscoi, , Course of meteorology, Leningrad, Gidrometeoizdat, (1962) (in Russian). [26] A.I. Rusanov and V.L. Kusmin, On the influence of electric field on the surface tension of the polar liquid, Kolloidnyi Journal, 39 (1977) 388 (in Russian). [27] A.I. Rusanov, To thermodynamics of nucleation on charged centers, Doklady Academii Nauk, USSR, 238 (1978) 831 (in Russian). [28] J.M. Meek and J. Craggs, Electrical Breakdown of Gases, Oxford, the Claredon Press (1953). [29] Ermakov , Molnii-sledy chastiz sverchvysokich energii, Nayka I zhizn, Moskva, Prosveshenie (1993) 92 (in Russian). [30] A.V. Gurevich , G.M. Molikh, and R.A. Roussel-Dupre, Runaway mechanism of air breakdown and preconditioning during a thunderstorm, Phys. Lett., A, 165 (1992) 463. [31] A.V. Gurevich, K.P. Zubin, R.A. Roussel-Dupre, Lightning initiation by simultaneous effect of runaway breakdown and cosmic ray showers, Phys. Lett., A, 254 (1999) 79. [32] S.V. Veretenenko and M.I. Pudovkin, Effects of forbush-decreases in cloudiness variations, Geomagn. and Aeronomy, 34 (1994) 38 (in Russian). [33] Y.I. Stozhkov , J. Zullo, Jr., I.M. Martin, G.Q. Pellegrino, H.S. Pinto, P.C. Bezerra, G.A. Bazilevskaya, V.S. Machmutov, N.S. Svirzevskii, A. Turtelli, Jr., Rainfalls during great Forbush-decreases, Nuovo Cimento, 18C (1995) 335. [34] Y.I. Stozhkov , P.E. Pokrevsky, J. Zullo, Jr., I.M. Martin, V.P. Ohlopkov, G.Q. Pellegrino, H.S. Pinto, P.C. Bezerra, A. Turtelli, Jr., Influence of charged particle fluxes on precipitation, Geomagn. and Aeronomy, 36 (1996) 211 (in Russian). [35] H. Svensmark and E. Friis-Christensen, Variation of cosmic ray flux and global coverage - a missing link in solar-climate relationships, Journ. Atmospheric and Solar-Terrest. Physics, 59 (1997) 1225. [36] R.E. Orville, G.R. Huffines, Lightning ground flash measurements over contiguous United States: a ten-year summary 1989-1998, Proc. 11th Intern. Conf. Atmospher. Electr., Alabama, USA, (1999) 412. [37] L.G. Kachurin, Fisicheskie osnovy vosdeistviya na atmosfernye processy, Gidrometeoisdat, Leningrad (1990) (in Russian). [38] New Scientist, New method of production artificial precipitation, 151 (1996) 10. CLOUD PROPERTY SURVEY FROM SATELLITE OBSERVATIONS USING VERTICAL SOUNDERS (TOVS PATH-B) AND IMAGERS (ISCCP) C. J. Stubenrauch and F. Eddounia Laboratoire de Météorologie Dynamique, Ecole Polytechnique, Palaiseau, France Abstract Since about 1980 the use of satellite radiometers allows a continuous survey of cloud properties over the whole globe. We compare the evolution of cloud amount and effective cloud amount from two different cloud climatologies : ISCCP (International Satellite Cloud Climate Project), using imagers onboard geostationary and polar orbiting weather satellites, and TOVS Path-B, obtained from TIROS-N Operational Vertical Sounder (TOVS) measurements onboard polar orbiters. Over a period of about ten years (1983-1995), the average cloud amount of about 67% and the effective cloud amount of about 53% are quite stable. The most important disturbance within this period was the volcanic eruption of Mount Pinatubo in June 1991. The slight increase of the visible reflectances by the volcanic aerosols has led to a slight overestimation of cloud optical thickness by ISCCP and hence to a slight overestimation of low clouds and underestimation of the amount of high clouds during the following year. Since infrared radiation is less affected by volcanic aerosols, TOVS cloud properties should be more reliable. The stable TOVS Path-B effective high cloud amount over the whole period in the tropics indicates that the volcanic aerosols do not change the properties of high clouds on this scale. 1. CLOUD OBSERVATIONS FROM SATELLITE Only satellite observations are capable to give a continuous survey of the state of the atmosphere over the whole globe. At present, twenty years of these measured radiances are available. In order to convert these radiances into cloud properties, complex inversion algorithms are necessary. These algorithms consist of two parts : i) cloud detection and ii) determination of cloud properties using radiative transfer models. In the following we present two global cloud climatologies. Whereas ISCCP is the cloud climatology with the best diurnal sampling and spatial resolution, TOVS Path-B cirrus (semi- transparent ice cloud) properties, obtained from vertical sounders with a relatively high spectral resolution, are especially reliable, day and night. However, one has to keep in mind that both climatologies give information only on the uppermost cloud. For climate studies, using one of these datasets, it is important to understand how cloud properties are perceived by these different instruments and inversion methods. A detailed comparison has shown that both datasets agree quite well [1]. Discrepancies can be explained by differences in temperature profiles, horizontal (partial cloud cover) and vertical (multi-layer clouds) heterogeneities. For example, in the case of thin cirrus overlying low clouds, one determines with TOVS the cirrus properties, whereas ISCCP determines a mixture of both clouds, from the visible channel. 1.1 ISCCP climatology For its global cloud climatology, the International Satellite Cloud Climatology Project (ISCCP) [2,3] puts emphasis on temporal and spatial resolution, rather than on spectral resolution, by using one visible (VIS, day only) and one infrared (IR) atmospheric window radiance measurement from imagers on the suite of geostationary and polar orbiting weather satellites. Time sampling is three hourly and the initial spatial resolution of about 5 km is sampled to intervals of about 30 km, which means that about one pixel out of 36 is kept for cloud information. The first ISCCP dataset has been thoroughly studied (e. g. [4]-[9]). Some of these studies have led to a recent re-analysis [3], mostly improving the treatment of cirrus and polar clouds. ISCCP has reprocessed eleven years of data (D- series) from 1983 until 1993. The processing has just taken up again; by the time of writing data are available until 1997, and the whole data period should be available within the next months. Clouds are detected through a variable IR-VIS threshold test which compares the measured radiances to ‘clear sky composite’ radiances that have been inferred from a series of statistical tests on the space and time variations of the IR and VIS radiances [10]. These clear sky conditions are associated with low IR and VIS spatial and temporal variability. ISCCP cloud properties are determined for each pixel by comparing the observed radiances with a detailed radiative transfer model. The model includes the effects of the atmosphere, with properties specified from the operational analysis of the TOVS data (only one profile per day is available), and surface determined from the clear radiances. Cloudy pixels are assumed to be covered completely by a single homogeneous cloud layer. Cloud-top temperature, Tcld, is first retrieved assuming that all clouds are black bodies. During daytime, when VIS radiances are available to retrieve cloud optical thicknesses, t, the cloud-top temperature of ‘transmissive’ clouds (t < 11) is corrected to account for the radiation transmitted from below. This means that Tcld is decreased as a function of t. In the first version of the ISCCP data (C-series), all clouds were represented in the radiative model by a cloud composed of 10 mm radius spherical liquid water droplets; in the new ISCCP dataset (D-series), clouds with Tcld ≥ 260 K are treated with the same liquid cloud model, but all clouds with Tcld < 260 K are treated with a model cloud composed of 30 mm ice polycrystals [3]. Tcld is converted into cloud-top pressure, pcld, using the operational TOVS atmospheric profiles. Clouds can be classified according to three pcld intervals (separated at 440 and 680 hPa). During daytime, clouds are classified into nine types, by separating each of the three cloud height categories into thin, medium and thick clouds according to three t intervals (divided at 3.6 and 23). Among many other variables, the D2 dataset gives statistics on cloud amount, CA, as calculated in Equation (1), and cloud amount separately for high, midlevel and lowlevel clouds at a spatial resolution of 2.5∞. CA = Ncld / Ntot. (1) wh ere Ncld is the number of cloudy pixels within the grid, and Ntot is the total number of pixels within the grid. 1.2 TOVS Path-B climatology The Improved Initialization Inversion (3I) algorithms [11] convert infrared and microwave radiation measured from the TIROS-N Operational Vertical Sounder (TOVS) onboard the NOAA polar orbiters into atmospheric temperature and humidity profiles and into cloud and surface properties. Within the framework of the NOAA/NASA Pathfinder Program, eight years of TOVS data (1987- 1995) have already been processed at LMD. This TOVS Path-B dataset provides these atmospheric parameters at a spatial resolution of 1∞ [12]. Results for the whole TOVS observation period from 1979 until now should be available by the end of 2001, since the re-calibration of the HIRS brightness temperatures obtained by comparing airmass-averaged brightness temperatures computed from radiosonde measurements to collocated observed brightness temperatures has just taken up again for the extending period. The 3I algorithms are based on i) the Thermodynamic Initial Guess Retrieval (TIGR) dataset, describing ~2000 different atmospheric conditions extracted from a huge collection of radiosonde measurements and ii) a fast line-by-line radiative transfer model, Automatized Atmospheric Absorption Atlas (4A, [13]), simulating clear sky and cloudy radiances at 30 pressure levels. Cloud det ect ion is performed at HIRS spatial resolution (~17 km at nadir) by eight (seven) threshold tests during daytime (nighttime). An important part of the cloud detection is the use of simultaneous MSU radiance measurements. Since the latter probe through the clouds, they are used to predict clear sky IR brightness temperatures which are compared to those of the HIRS instrument for all individual pixels to decide if they are cloudy. A summary of the 3I cloud detection scheme is given in Table 1 of [14]. To insure more coherence with the MSU spatial resolution (~100 km at nadir), the HIRS radiances are averaged separately over clear pixels and over cloudy pixels within 100 km x 100 km regions. Cloud properties are determined from the averaged cloudy pixel radiances assuming that all cloudy pixels are covered by a single homogeneous cloud layer. The average cloud-top pressure, pcld, and the average effective cloud amount over cloudy pixels, Necld, are obtained from four radiances in the 14 mm CO2-absorption band (with peak responses from 400 to 900 hPa levels in the atmosphere) 2 and one in the 11 mm IR atmospheric win dow by minimizing a weighted c [15]. Empirical weights reflect the effect of the brightness temperature uncertainty within a TIGR airmass class on these radiances at the various cloud levels. The method is based on the coherence of the effective cloud emissivity, Necld, in Equation (2), obtained from the five wavelengths at the pressure level of the real cloud. IImi()ll- clr () i Npeelcld() cld@= Np cld (,) k i for i = 4,8 (2) Ipcld(,) kll i- I clr () i wh ere li is the wavelength of HIRS channel i, pk is the pressure level k out of 30 levels, Im is the measured radiance, Iclr is the retrieved clear sky radiance, and Icld is the calculated radiance emitted by a homogeneous opaque single cloud layer. Tcld is obtained from pcld using the retrieved 3I atmospheric temperature profiles. The cloud amount, CA, is again determined as in Equation (1), but this time in each 1∞ grid. The effective cloud amount over a 1∞ grid, eN, is the product of cloud amount and effective cloud emissivity, Necld: eN = CA x Necld (3) Cloud types are defined by the cloud-top pressure and effective cloud amount. High clouds (pcld < 440 hPa) are divided into three categories: opaque (Necld > 90%), cirrus (90% < Necld < 50%) or thin cirrus (Necld < 50%). Since midlevel (440 h Pa < pcld < 680 hPa) and low-level (pcld > 680 hPa) clouds have a smaller horizontal extension, only two classes in each height category can be distinguished: mostly cloudy or overcast (eN > 50%) and partly cloudy (eN < 50%) fields. Their relatively high spectral resolution make infrared sounders very useful for the determination of cloud properties (frequency, altitude, cloud top temperature and effective emissivity), day and night. Their coarse spatial resolution (20 km) has less effect on clouds with large spatial extents like cirrus clouds. In addition to cloud height and effective emissivity, we start to retrieve mean effective ice crystal sizes for cirrus clouds, taking advantage of the fact that spectral cirrus emissivity differences between 8 and 11 mm depend on this parameter [16]. An eight year survey of these cirrus properties will be available within the framework of the European project CIRAMOSA (CIrrus microphysical properties and their effect on RAdiation: survey and integration into climate MOdels using combined SAtellite observations ; web-site : http ://www.lmd.polytechnique.fr/CIRAMOSA/Welcome.html). 2. CLOUD CLIMATE STUDIES The decade of retrieved cloud data is certainly not yet enough to study climate change, but these datasets give a good starting point to study cloud properties and their variations in correlation with natural events like volcanic eruptions or the El Niño event and La Niña event. Whereas in normal conditions the trade winds blow towards the west across the Tropical Pacific and therefore pile up warm surface water in the west Pacific, during El Niño the trade winds relax in the central and western Pacific leading to a penetration of warm water towards the east. El Niño and La Niña are opposite phases of the El Niño-Southern Oscillation (ENSO) cycle, with El Niña sometimes referred to as the cold phase of ENSO and El Niño as the warm phase of ENSO. Within the period from 1983 to 1995, there were one El Niña event (1985) and a following El Niño event (1986/87) and then a rapid succession of El Niño events in the 90's (1991/92, 1993 and 1994). The eruption of Mount Pinatubo in the Philippines in June 1991 has spread a huge amount of sulfate aerosols into the stratosphere which staid in the atmosphere for more than two years [17]. Recently, a link between the variation of galactic cosmic rays intensity and cloud amount on earth has been found by Svensmark et al. [18], which could not be confirmed by Kristjansson et al. [19]. 2.1 Cloud amount variation In order to give a first impression of cloud amount and its variations with time and seasons during the period from 1983 until 1995, we present in Fig. 1 monthly averages of cloud amount from ISCCP (D2) and TOVS Path-B ('3I') compared to the TOVS Path-B effective cloud amount. The ISCCP D2 cloud amount is obtained during daytime from IR and VIS data, and the cloud amount during nighttime, using only IR data, is adjusted to the daytime results after a comparison of both methods during daytime [20]. TOVS Path-B data for this analysis are used only from the NOAA- 10 and NOAA-12 satellites with local observation times at 7h30 am and pm. The NOAA-11 satellite which was launched in 1988 for a local observation time of 1h30 am and pm, has drifted strongly during its operation with a local observation time of 5h30 in 1991. This drift has strong consequences on the diurnal sampling of the data, especially over land where the diurnal cycle of clouds can be strong [7]. In Fig. 1 we observe average global cloud amounts of 67% (ISCCP) and 77% (TOVS Path-B), whereas the effective cloud amount, taking into account also the cloud opacity, is only 53%. The 10% difference between ISCCP and TOVS Path-B cloud amount can be explained 1) by a higher sensitivity to thin cirrus clouds of TOVS due to its better spectral resolution and 2) by the larger HIRS pixel size for which it can be declared cloudy even if it is only half covered by clouds [14]. Cloud amount alone is not a sufficient variable to look for climate changes. One also should look at the thickness of clouds, which is possible only during day with ISCCP. On the other hand, the TOVS Path-B effective cloud amount which is reliable day and night, combines cloud cover and cloud thickness. Radiative effects of clouds depend on effective cloud amount and cloud height. The inversion method also takes the smaller measured radiances of partly covered pixels into consideration. We observe a seasonal cycle in cloud amount with a maximum in northern hemisphere winter and a minimum in summer. A minimum of effective cloud amount in northern hemisphere winter leads to the assumption of a larger amount of thin cirrus clouds during this season. Within a few percent these global cloud properties seem stable during this whole period. globe 0.85 3I effective 0.8 ISCCP 3I 0.75 0.7 0.65 0.6 cloud amount 0.55 0.5 0.45 84 86 88 90 92 94 year Fig. 1. Monthly mean ISCCP cloud amount, TOVS Path-B (3I) cloud amount and effective cloud amount as a function of time. 2.2 Variation of low, midlevel and high clouds In this section we explore cloud amount and effective cloud amount separately for low, midlevel and high clouds, as defined in sections 1.1 and 1.2. We compare in Figs. 2a to 2c monthly mean cloud amount over the globe for these cloud types from ISCCP using IR and VIS radiances ('ISCCP day') and using IR radiances only ('ISCCP IR', day and night). The latter data have been used by Marsh and Svensmark [21] to reveal a correlation between galactic cosmic ray intensity and low cloud amount. TOVS Path-B effective cloud amount ('3I') for these cloud types, which is defined as their frequency times effective cloud amount during appearance, is also shown in Figs. 2a to 2c. For low and midlevel clouds, we notice a slightly larger cloud amount from the 'ISCCP IR' analysis than from the ISCCP day measurements. However, for high clouds, the 'ISCCP IR' analysis yields a 8% smaller cloud amount than the 'ISCCP day' analysis. As the IR analysis converts the IR brightness temperature into cloud-top temperature under the hypothesis that clouds are black bodies, the cloud-top temperature of semi-transparent clouds is overestimated corresponding to an underestimation of their height. Therefore, the 'ISCCP IR' low and midlevel cloud amounts contain in addition also higher, semi-transparent clouds, whereas the 'ISCCP IR' high cloud amount contains only high opaque clouds. A separation into cloud height does only make sense when using the more reliable 'ISCCP day' analysis. globe 35 30 25 20 3I effective low cloud amount 15 ISCCP day ISCCP IR 10 84 86 88 90 92 94 year globe 35 3I effective ISCCP day 30 ISCCP IR 25 20 mid cloud amount 15 10 84 86 88 90 92 94 year globe 35 3I effective ISCCP day 30 ISCCP IR 25 20 high cloud amount 15 10 84 86 88 90 92 94 year Fig. 2. Monthly mean ISCCP cloud amount from VIS and IR radiances (day) and from IR radiances only (IR), as well as TOVS Path-B (3I) effective cloud amount as a function of time a) for low clouds, b) for midlevel clouds and c) for high clouds. The difference between ‘ISCCP day’ cloud amount and TOVS Path-B effective cloud amount should give an estimation of cloud thickness. This difference is negligible for low clouds in northern hemisphere winter, but 5% for low clouds in northern hemisphere summer. Also midlevel clouds and high clouds have in general an about 6% smaller effective cloud amount. This means that on average low clouds are slightly thicker in northern hemisphere winter, and there is a high occurrence of semi-transparent cirrus clouds over the globe (about 25%, not shown). The slight increase of ‘ISCCP day’ low cloud amount and TOVS Path-B effective low cloud amount and decrease of ‘ISCCP day’ high cloud amount are related to the volcanic eruption of Mount Pinatubo as we will show in the following section. 2.3 Evolution of cloud amount In order to look for systematic changes in cloud amount, we analyze running means over twelve months which take out seasonal variations and subtract them from the average over the whole data period (1983-1993 for ISCCP and 1987-1995 for TOVS Path-B). Figs. 3a to 3d show the evolution of ISCCP cloud amount and TOVS Path-B effective cloud amount ('3I') a) over the globe, b) in the tropics (15∞N to 15∞S), c) over northern hemisphere midlatitudes (30∞N to 60∞N) and d) over southern hemisphere midlatitudes (30∞S to 60∞S). globe tropics 2 3I effective 4 3I effective ISCCP 3 ISCCP 1 2 1 0 0 -1 -1 cloud amount cloud amount -2 -3 -2 -4 84 86 88 90 92 94 84 86 88 90 92 94 year year NH mid latitude SH mid latitude 2 3I effective 2 3I effective ISCCP ISCCP 1 1 0 0 -1 -1 cloud amount cloud amount -2 -2 84 86 88 90 92 94 84 86 88 90 92 94 year year Fig. 3. Difference between running mean over twelve months of ISCCP cloud amount and TOVS Path-B (3I) effective cloud amount and mean over the whole data period (1983-1993 for ISCCP and 1987-1995 for TOVS Path-B) as a function of time. Each point is plotted against the seventh month of each one year period. Results are shown a) over the globe, b) in the tropics, c) over northern hemisphere midlatitudes and d) over southern hemisphere midlatitudes. Over the globe, variations are within 1%. Differences in behavior between ISCCP cloud amount and TOVS Path-B effective cloud amount appear after the volcanic eruption of Mount Pinatubo in 1991. This appears even clearer by looking at the tropics where the TOVS Path-B effective cloud amount increased by 4% and decreases only a year later. The maxima of cloud amount in 1987 is probably related to the El Niño event [3]. This effect seems to be the highest in the northern hemisphere midlatitudes. Normally, the El Niño event shifts the tropical convection from West to East, but perhaps the convection also shifts slightly northwards. This has to be studied more in detail by studying geographical maps. 2.4 Evolution of high and low clouds If one wants to study correlations between cloud amount and the variation of galactic cosmic ray intensity, one should look for them at high clouds and at higher latitudes, since the intensity decreases by entering the atmosphere and since the Earth magnetic field is lower at these latitudes. Figs. 4a to 4d shows the evolution of ISCCP high cloud amount from ISCCP using IR and VIS radiances ('ISCCP day') and using only IR radiances ('ISCCP IR', day and night) and of TOVS Path-B effective high cloud amount ('3I') a) over the globe, b) in the tropics, c) over northern hemisphere midlatitudes and d) over southern hemisphere midlatitudes. globe tropics 3 4 3I effective 3I effective 3 2 ISCCP IR ISCCP IR ISCCP day 2 ISCCP day 1 1 0 0 -1 -1 -2 high cloud amount -2 high cloud amount -3 -4 -3 84 86 88 90 92 94 84 86 88 90 92 94 year year NH mid latitude SH mid latitude 3 3 3I effective 3I effective 2 ISCCP IR 2 ISCCP IR ISCCP day ISCCP day 1 1 0 0 -1 -1 high cloud amount -2 high cloud amount -2 -3 -3 84 86 88 90 92 94 84 86 88 90 92 94 year year Fig. 4. Difference between running mean over twelve months of ISCCP high cloud amount from VIS and IR radiances (day) and from IR radiances only (IR), as well as TOVS Path-B (3I) effective high cloud amount and mean over the whole corresponding data period as a function of time. Each point is plotted against the seventh month of each one year period. Results are shown a) over the globe, b) in the tropics, c) over northern hemisphere midlatitudes and d) over southern hemisphere midlatitudes. From these figures we conclude that there are no significant variations of the effective high cloud amount and of the 'ISCCP IR' high cloud amount, corresponding to high opaque clouds only. However, the 'ISCCP day' high cloud amount decreases by 2.5% globally and 4% in the tropics after the volcanic eruption of Mount Pinatubo in 1991. This effect can be explained by the increase of the VIS reflectance by the volcanic aerosols of this eruption which have had an optical thickness of more than 0.1. In the ISCCP cloud property retrieval this increase in VIS reflectance affects then the optical thickness of the clouds, and an overestimation of optical thickness leads then to an underestimation of cloud height [3]. This can be seen in Figs. 5a and 5b which show the evolution ISCCP low cloud amount from ISCCP using IR and VIS radiances ('ISCCP day') and using only IR radiances and of TOVS Path-B effective low cloud amount ('3I') a) over the globe and b) in the tropics. The low cloud amount increase from 'ISCCP day' by 3% in the tropics (and midlevel cloud amount increase by 1.5%, not shown) shows this effect clearly, whereas the 'ISCCP IR' low cloud amount, containing a mixture of low and semi-transparent higher clouds over the whole period, does not show such an increase. Therefore, the increase of the TOVS Path-B effective low cloud amount which is nearly identical to the increase of the 'ISCCP day' low cloud amount should also be slightly overestimated. Whereas the TOVS Path-B cloud properties should in general not be affected by the volcanic aerosols since they are retrieved from IR radiances, there is one cloud test out of eight during day which makes use of the VIS reflectance [14]. This test should be affected by the volcanic aerosols, and since low clouds show a stronger contrast with the surface in albedo than in temperature, the effective cloud amount of low clouds should be more affected. Also, the threshold is lower over ocean (15%) than over land (20%). Therefore, cloud amount over ocean should be more affected. This effect has been analyzed by separating ocean and land and NOAA- 10/NOAA-12 observations and NOAA-11 observations. globe tropics 2 3I effective 4 3I effective ISCCP IR 3 ISCCP IR ISCCP day ISCCP day 1 2 0 1 0 -1 -1 low cloud amount -2 low cloud amount -2 -3 84 86 88 90 92 94 84 86 88 90 92 94 year year NH mid latitude SH mid latitude 2 3I effective 2 3I effective ISCCP IR ISCCP IR ISCCP day ISCCP day 1 1 0 0 -1 -1 low cloud amount low cloud amount -2 -2 84 86 88 90 92 94 84 86 88 90 92 94 year year Fig. 5. Difference between running mean over twelve months of ISCCP low cloud amount from VIS and IR radiances (day) and from IR radiances only (IR), as well as TOVS Path-B (3I) effective low cloud amount and mean over the whole corresponding data period as a function of time. Each point is plotted against the seventh month of each one year period. Results are shown a) over the globe, b) in the tropics, c) over northern hemisphere midlatitudes and d) over southern hemisphere midlatitudes. Figs. 6 show the evolution of these different TOVS Path-B cloud amounts a) over ocean and b) over land. Indeed, over ocean, where there is no effect with NOAA-11 observations at 1h30 am, taken during nighttime, one observes a strong cloud amount increase for the NOAA- 10/NOAA-12 observations which are taken when the sun has a low zenith angle, already a difficult time for analyzing VIS reflectances. Over land, there is no effect, with exception at 7h30 pm where a much smaller cloud amount overestimation than over ocean can be seen. However, due to the basis of cloud emissivity coherence in our retrieval method the strong overestimation in cloud amount is nearly compensated by a smaller retrieved effective cloud emissivity, so that the final effect for low clouds is only about 3% in the tropics. As we have seen in Figs. 4, effective high cloud amount is not affected by volcanic aerosols. Nevertheless, we will improve this cloud detection test in the next TOVS Path-B re-analysis. tropical ocean tropical land 100 100 NOAA1012 am NOAA1012 am 95 NOAA11 am 95 NOAA11 am 90 NOAA1012 pm 90 NOAA1012 pm NOAA11 pm NOAA11 pm 85 85 80 80 75 75 cloud amount 70 cloud amount 70 65 65 60 60 88 89 90 91 92 93 94 95 88 89 90 91 92 93 94 95 year year Fig. 6. Running mean over TOVS Path-B (3I) cloud amount from different satellite observations as a function of time. Each point is plotted against the seventh month of each one year period. Results are shown for a) tropical ocean and b) tropical land. 3. CONCLUSION AND OUTLOOK Satellite observations provide a unique possibility to survey cloud properties over a long period of time. During the observed decade (1983-1995), the average cloud amount of about 67% and effective cloud amount of about 53% are stable within 2% over the globe. Within this period, the most important disturbance was the volcanic eruption of Mount Pinatubo in the Philippines in June 1991. The volcanic aerosols which had an optical thickness of more than 0.1 slightly increased the VIS reflectances. Whereas the ISCCP cloud amount and TOVS Path-B effective high cloud amount were not affected by this event, the 'ISCCP day' high cloud amount is slightly underestimated (4.5% in the tropics and 2.5% over the globe) and the 'ISCCP day' low amount overestimated (4% in the tropics and 1% over the globe). Nevertheless, a separation into cloud height does only make sense when using the more reliable 'ISCCP day' analysis, since the 'ISCCP IR' low and midlevel cloud amounts contain in addition also higher, semi-transparent clouds, whereas the 'ISCCP IR' high cloud amount contains only high opaque clouds. The TOVS Path-B effective low cloud amount is also slightly overestimated (4% in the tropics and 1% over the globe), because one of the cloud detection tests makes use of the VIS reflectance. As has been pointed out earlier, if there are any correlations between the galactic cosmic ray intensity and cloud properties these should occur for high clouds and at higher latitudes. The TOVS Path-B effective high cloud amount which is stable within 1% over the whole observation period does not show any correlation with the cosmic ray intensity variations. It should be noted that cloud radiative effects are determined not only by cloud amount but depend also on cloud thickness and effective cloud amount includes both variables. Both datasets will be extended. By the time of writing, ISCCP data processing has already started again, and the whole data period (1983 until now) should be available within the next months. TOVS Path-B processing just started with the re-calibration of the HIRS brightness temperatures, and the whole TOVS observation period from 1979 until now should be available by the end of 2001. ACKNOWLEDGEMENTS Our thanks to W. B. Rossow for many stimulating discussions within the last years in order to advance in understanding both cloud datasets. We also want to thank the rest of our ARA (Analyse du Rayonnement Atmosphérique) group for their support, and especially S. Serrar and R. Armante for their help in computational matters and G. Rädel for bringing our attention to CERN again. REFERENCES [1] Stubenrauch, C. J., W. B. Rossow , N. A. Scott, and A. Chédin, Clouds as Seen by Satellite Sounders (3I) and Imagers (ISCCP): III) Spatial Heterogeneity and Radiative Effects, J. Climate 12 (1999) 3419-3442. [2] W. B. Rossow and R. A. Schiffer, ISCCP Cloud Data Products. Bull. Amer. Meteor. Soc. 72 (1991) 1-20. [3] W. B. Rossow and R. A. Schiffer, Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc. 80 (1999) 2261-2287. [4] R. Fu, A. D. Del Genio, and W. B. Rossow, Behavior of deep convective clouds in the tropical Pacific deduced from ISCCP radiances. J. Climate 3 (1990) 1129-1152. [5] G. Tselioudis, W. B. Rossow, and D. Rind, Global patterns of cloud optical thickness variation with temperature, J. Climate 5 (1992) 1484-1495. [6] S. A. Klein and D. L. Hartmann, The seasonal cycle of low stratiform clouds. J. Climate 6 (1993) 1587-1606. [7] B. Cairns, Diurnal variations of cloud from ISCCP data. Atm. Res. 37 (1995) 133-146. [8] X. Liao, W. B. Rossow and D. Rind,: Comparison between SAGE II and ISCCP High-Level Cloulds. Part II: Locating Cloud Tops. J. Geophys. Res. 100 (1995) 1137-1147. [9] Y. Jin, W. B. Rossow and D. P. Wylie, Comparison of the Climatologies of High-Level Clouds from HIRS and ISCCP. J. Climate. 9 (1996) 2850-2879. [10] W. B. Rossow and L. C. Garder, Cloud detection using satellite measurements of infrared and visible radiances for ISCCP. J. Climate 6 (1993) 2341-2369. [11] A. Chédin, N. A. Scott, C. Wahiche and P. Moulinier, The Improved Initialized Inversion method: A high resolution physical method for temperature retrievals from the TIROS-N Series. J. Clim. Appl. Meteor. 24 (1985) 124-143. [12] N. A. Scott, A. Chédin, R. Armante, J. Francis, C. J. Stubenrauch, J.-P. Chaboureau, F. Chevallier, C. Claud and F. Chéruy, Characteristics of the TOVS Pathfinder Path-B Dataset. Bull. Amer. Meteor. Soc. 80 (1999) 2679-2701. [13] N. A. Scott, and A. Chédin, A fast line-by-line method for atmospheric absorption computations: The Automized Atmospheric Absorption Atlas. J. Appl. Meteor. 20 (1981) 802-812. [14] C. J. Stubenrauch, W. B. Rossow , F. Chéruy, N. A. Scott and A. Chédin, Clouds as Seen by Satellite Sounders (3I) and Imagers (ISCCP): I) Evaluation of Cloud Parameters, J. Climate 12 (1999) 2189-2213. [15] C. J. Stubenrauch, A. Chédin, R. Armante and N. A. Scott, Clouds as Seen by Satellite Sounders (3I) and Imagers (ISCCP): II) A New Approach for Cloud Parameter Determination in the 3I Algorithms, J. Climate 12 (1999) 2214-2223. [16] C. J. Stubenrauch, R. Holz, A. Chédin, D. Mitchell and A. J. Baran, Retrieval of Cirrus Ice Crystal Sizes from 8.3 and 11.1 mm Emissivities Determined by the Improved Initialization Inversion of TIROS-N Operational Vertical Sounder Observations, J. Geophys. Res. 104 (1999) 31793-31808. [17] M. P. McCormick, L. W. Thomason, and C. R. Trepte, Atmospheric effects of the Mt Pinatubo eruption, Nature 373 (1995) 399-404. [18] H. Svensmark and E. Friis-Christensen, Variation of cosmic ray flux and global cloud coverage: A missing link in solar climate relationships, J. Atmos. Sol. Terr. Phys. 59 (1997) 1225-1232. [19] J. E. Kristjansson, and J. Kristiansen, Is there a cosmic ray signal in recent variations in global cloudiness and cloud radiative forcing?, J. Geophys. Res. 105 (2000) 11,851- 11,863. [20] W. B. Rossow, A. W. Walker, D. Beuschel, and M. Roiter, International Satellite Cloud Climatology Project (ISCCP) description of new cloud datasets, World Climate Research Programme (ICSU and WMO) WMO/TD-No.737 (1996) 115pp. [21] N. Marsh, and H. Svensmark, Cosmic rays, clouds, and climate, Space Science Reviews 94 (2000) 215-230. ATMOSPHERIC ELECTRICITY AND CLOUD MICROPHYSICS R. G. Harrison Department of Meteorology, The University of Reading, P.O. Box 243, Reading RG6 6BB, UK Abstract The terrestrial atmospheric electrical system covers a range of dimensional scales, from charged molecular clusters to convective cloud systems. Charge-exchange associated with thunderclouds leads to positive charge in the upper conductive regions of the atmosphere and a net negative charge on the planetary surface. In non-thunderstorm regions, a vertical ionic current flows, replenishing the air with molecular ions otherwise removed by attachment, recombination or nucleation processes. Ions may have indirect effects on non-thunderstorm clouds, and therefore conceivably on climate, via cloud microphysical processes. Cloud Condensation Nuclei (CCN) and Ice Nuclei (IN) are necessary for the formation of water clouds and freezing of ice clouds respectively. In both cases, ionisation may be important: it is now known that ultrafine aerosol can be formed from ionisation, probably providing an additional source of CCN. It is also known that electrified aerosol, perhaps active as IN, can be collected by droplets more effectively than neutral particles. 1. INTRODUCTION Electrical processes in atmospheric air arise from the combined effect of natural ionisation and the natural electric fields generated indirectly by charge separation in thunderclouds. In non- thunderstorm regions, which probably constitute the majority of the global cloud area, the electrical processes will not generate the large breakdown electric fields associated with lightning, but microscopic aerosol particles acquire charges by diffusion of the molecular cluster ions formed from ionisation. In this overview, the effect of small charges on aerosol particles and droplets are considered. Since the charge arises from radiolysis of air by cosmic rays and natural radioactivity, the discussion here is structured in terms of the processes associated with charge generation and removal, including tutorial material on microphysical cloud processes. It has been observed (Marsh and Svensmark, 2000) that there is a correlation between low cloud properties and the neutrons produced by cosmic rays. 2. THUNDERSTORMS AND GLOBAL ELECTRIFICATION The atmospheric electrical system originally discussed by Wilson (1929) can be simplified into an electric circuit in which thunderstorms separate charge in convective regions. The charge separation leads to a potential difference between conductive regions of the upper atmosphere and the surface, which causes an ionic leakage current to flow vertically (figure 1). Currents of order 2000A flow in the circuit, with an upper atmosphere potential of ~300kV. The conduction current density in undisturbed regions is ~2pA.m-2. The charge-exchange processes within thunderclouds are complicated, and probably result from the interaction between rising ice crystals rising and riming soft hail (graupel). Typical microphysical collisions exchange charge with typical magnitudes of tens of femtoCoulombs, but the precise magnitude and polarity is greatly influenced by the liquid water content and temperature (MacGorman and Rust, 1998). Figure 1. The global atmospheric electrical circuit (from Harrison, 1997). 3. ATMOSPHERIC PROPERTIES AND CLOUD MICROPHYSICS 3.1 Atmospheric properties 3.1.1 Bulk properties The troposphere (lower atmosphere) shows variations in temperature and water content, and partitioning of the water concentration between liquid, solid or vapour forms is critical to the formation and distribution of clouds. Figure 2 shows a vertical sounding of temperature and humidity, which illustrates the atmospheric structure. The presence of low cloud (which was observed from the surface) is evident from the sharp increase in relative humidity, marked as A. B shows a slight temperature inversion associated with the top of the planetary boundary layer, and at C the temperature ceases to fall with height, at the tropopause. It is clear that there is considerable variability in the relative humidity during the ascent, and in the region where cloud was identified optically. Radiosonde ascent on 22nd March 1998 at 1400h, 1000 (launched from Medina Valley, Isle of Wight, UK) 80 pressure 60 800 air temperature 40 relative humidity Relative Humidity(%) 20 600 temperature (degC) and 0 C 400 A -20 -40 200 B -60 pressure /mbar altitude/m 0 -80 0 5000 10000 15000 20000 Figure 2. Vertical atmospheric sounding in non-frontal synoptic conditions, showing relative humidity, temperature and pressure variations with height. (A, B and C are discussed in the text.) 3.1.2 Microphysical properties In addition to variability in temperature and humidity, there is a considerable variety in the sizes and abundance of aerosol particles and cloud droplets present in the atmosphere. Figure 3 shows a comparison of the sizes of cloud droplets, raindrops, and a condensation nucleus. The typical molecular cluster comprising an atmospheric small ion will have a diameter less than one nanometre. Figure 3. Size spectrum of particles present in a typical cloud. (from Rogers and Yau, 1989). 3.2 Cloud Microphysics The concentration of water vapour in air can be determined by its gaseous partial pressure, and, at any given temperature there is an associated maximum value of partial pressure due to water vapour, the saturation vapour pressure. Air containing sufficient water vapour to generate the saturation vapour pressure is saturated, with a relative humidity of 100%. Slightly greater relative humidities (supersaturations) can occur in localised regions, but they are never greater than a few percent, because of the abundance of aerosol particles on which the water can condense. Many different kinds of aerosol particles are capable of acting as condensation nuclei. Below 0¡C however, liquid water droplets may persist without freezing, although 0¡C is the temperature at which ice melts. Any liquid water droplet with a temperature below 0¡C is supercooled, in a thermodynamically unstable state in which freezing may be readily initiated by heterogeneous or homogeneous nucleation. In heterogeneous nucleation, the supercooled water freezes as a result of the presence of a suitable ice nucleus. Homogeneous nucleation occurs if cooling is continued further, and all supercooled water in atmospheric clouds becomes ice at temperatures colder than - 40¡C by this process. 3.2.1 Saturation vapour pressure, temperature and relative humidity At any given temperature T, the maximum partial pressure of water vapour, the saturation vapour pressure es(T) is given by the Clausius-Clapeyron equation as 1 des l = 2 (1) es dT RTv where l is the latent heat of vaporisation of water, and Rv is the gas constant for water vapour -1 -1 (461.5 J.kg .K ). es(T) can in principle be found by integration from equation (1), but l is also a function of temperature, which leads to many empirical formulae for es(T). Common forms include the exponential (Magnus) equation e.g. es(T)=6.112 exp [17.67 T / (243.5 + T) ] (2) where es is given in millibars and T is in Celsius. The relative humidity is the actual vapour pressure expressed as a fraction of es, at the same temperature. Supersaturation is expressed either as a percentage relative humidity greater than 100%, or as a saturation ratio S. (101% RH = 1% supersaturation = saturation ratio S =1.01). 3.2.2 Activation of condensation nuclei In the troposphere, supersaturations are never greater than a few percent, and are typically rather less. Consequently direct condensation onto ions, which permits visualisation of particle tracks in a Cloud Chamber (S ~ 4), cannot occur in the lower atmosphere. Condensation on aerosol particles, which are larger, does occur, however, and the minimum size of particle necessary depends on the degree of supersaturation. All aerosol particles are therefore potentially able to act as condensation nuclei (CN), if the supersaturation is sufficiently large, but it is the subset of particles able to cause condensation at atmospheric supersaturations which is of interest in cloud physics. These condensation nuclei are known as Cloud Condensation Nuclei (CCN). The vapour pressure over the curved water surface of a particle of radius r, es(r) is greater than es over a plane surface at the same temperature. If condensation occurs on a particle, its growth rate is proportional to the difference in between the bulk vapour pressure e and es(r). For e - es(r) > 0 the cloud droplet grows. This situation is rather more complicated in a mixed-phase cloud due to the differences in vapour pressure over ice and supercooled water. Ice particles grow at the expense of supercooled water in a mixed-phase system. shrinks grows saturation ratio S È b ˘ Êaˆ S(r) = 1- exp Í 3 ˙ Á ˜ Î r ˚ Ë r¯ 0.6% activated at 0.13µm ultrafines and ions Figure 4. Activation of particles at typical atmospheric saturation ratios. The maximum in the saturation ratio curve S(r) defines the minimum radius of particle required to act as a nucleus on which a cloud droplet to grow. For a supersaturation of 0.6%, a 0.13µm radius particle is required. A droplet smaller than this will evaporate. The function S(r) principally depends on a “curvature” term a, and a “solution” (dissolved salt) term b. (after Rogers and Yau, 1989). 3.2.3 Supercooling and ice nucleation Supercooled droplets are common under atmospheric conditions, and result from water droplets cooling in the absence of suitable ice nuclei (IN) to permit heterogeneous ice nucleation. At temperatures cooler than -40¼C all supercooled droplets begin to freeze by homogeneous nucleation. Only very few atmospheric aerosol particles can act as IN, typically less than 1% although the exact fraction increases as the droplets become colder. The ability of a particle to act as an ice nucleus depends on a variety of physical properties including its shape, solubility, crystal structure and its history in cloud processing. At warmer temperatures (-6 to -10¼C) ice multiplication occurs by mechanical production of ice splinters on freezing, generating additional ice fragments which are also able to act as IN. Figure 5 shows the dramatic temperature change occurring when a supercooled water droplet freezes, releasing latent heat. Temperature within 1µL water drop during supercooling and freezing 5.00 0.00 Td (drop) Te (environment) -5.00 -10.00 mperature / degC te -15.00 -20.00 700 750 800 850 time (from beginning of cooling) /seconds Figure 5. Time series of temperature Td within a supercooled water droplet, in an environment at temperature Te, as it freezes and releases latent heat (from Harrison and Lodge, 1998). 4. ATMOSPHERIC IONISATION AND ELECTRIFICATION 4.1 Steady-state ion concentrations Ion-pairs are continually produced in the atmosphere by radiolysis of air molecules, figure 6. The ions produced are rarely single species but clusters of water molecules around a central ion. Typical atmospheric ion concentrations in unpolluted air and fine weather are about 500 ions.cm-3 (Chalmers, 1967). There are three principal sources of high-energy particles which cause radiolysis: Radon isotopes, cosmic rays and terrestrial gamma radiation. The partitioning between the sources varies vertically. Near the surface, ionisation from turbulent transport of radon and other radioactive isotopes is important, together with gamma radiation from isotopes below the surface. Ionisation from cosmic rays is always present, comprising about 20% of the ionisation at the surface. The cosmic fraction increases with increasing height in the atmosphere and dominates above the planetary boundary layer. Figure 6. Formation of small ions by radiolysis of air molecules. Small ions consist of clusters of water molecules collected around a singly charged ion. They have a lifetime of the order of a hundred seconds. Clusters such as H3O+(H2O)n, H+(H2O)n, NO+(H2O)n and NO2+(H2O)n are common for the positive ions and O2-(H2O)n, CO4-(H2O)n, NO-(H2O)n or NO2-(H2O)n for the negative ions (Volland, 1984). The chemical difference between the species in the positive and negative ions leads to some physical asymmetries in the ion properties, with the negative ions more mobile. The ratio of mobilities m-/m+ ~ 1.2. 4.2 Ion balance equation Atmospheric small ions of both signs with number concentrations n+ and n- are governed by • • dn± =-qnnnabm - () aNada () (3) dt ±±Ú Â ±1,jj a =0 j =-• where the ions are produced at a rate q per unit volume. Ions (which are assumed to carry unit charges) are removed by ion-ion recombination (with recombination coefficient a), and by attachment to aerosol particles, which causes charge transfer to the aerosol. The aerosol attachment rate b±1,j(a) depends on aerosol particle radius a and the number of elementary charges j present on the aerosol particle of radius a (Gunn, 1954). In equation (3), the size and charge distributions of atmospheric aerosol particles are accounted for by the integral of number concentration N(r) over all particle radii, and by a sum across all possible particle charges at each radius. Recombination is the principal loss mechanism of ions in clean, aerosol free air. If aerosol is present, then ions are also lost by aerosol attachment. It is instructive to simplify the ion balance equation by neglecting the ion sign (i.e. n+ ª n- = n) and replacing the aerosol particle size distribution by an equivalent monodisperse particle number concentration Z. The ion-aerosol equation can then be written as dn =-qnnZab2 - (4) dt 4.2.1 Time dependent solution Integrating this equation gives the ion concentration n as a function of time t, for a zero initial ion concentration at time zero, as ÈÊ -+()ba22Zqt4 ˆ ˘ 22 - []-+()babZqZ4 - ÍË1 e ¯ ˙ nt()= Í ˙ (5) a Ê -+()ba22Zqt4 ˆ 2 Í 1 + e ˙ ÎË ¯ ˚ which highlights two interesting points. Firstly, if the ion-pair production rate q is uniform and the removal rates are also steady, the ion concentration tends to a steady value for large values of t. Secondly the equation can be simplified according to the situations in which attachment or recombination dominates as the removal mechanisms, according to whether an2 or nbZ is the bigger term. In the atmosphere in polluted air, these terms are roughly comparable, and therefore all the terms in equation (5) have to be evaluated. 4.2.2 Recombination Limit In the case of ion loss solely by recombination, such as in relatively aerosol-free regions of the atmosphere, equation (5) reduces to È 1- e-2 aqt ˘ È q ˘ Í()˙ nt()= Í ˙ (6) Îa ˚ Í 1 + e-2 aqt ˙ Î()˚ 1/2 and the steady-state concentration after a long time has elapsed is given by n• = (q/a) . Inserting typical atmospheric values of q ª10 ion-pairs cm-3 s-1and a =1.6 x 10-6 cm3 s-1 gives -3 n• = 2500 ion-pairs cm . Typical values of small ion concentrations observed in mountain air are about 500 ions cm-3 of each sign, suggesting that attachment processes are almost always significant in modulating the ion concentrations in the lower troposphere. 4.3 Aerosol electrification Collisions between the ions and atmospheric aerosol lead to charge-exchange and electrification of the aerosol, and the ion asymmetry ensures that the collisions do not lead to an average charge of zero. Local electric fields can cause further asymmetries, by depletion of one sign of ion concentration, and consequently substantial aerosol electrification can occur in such regions. The number concentration Nj of monodisperse aerosol particles carrying j elementary charges is given by the Modified Boltzmann Distribution (Clement and Harrison, 1992), as N j È 2 ˘ È 22 ˘ j Èn++m ˘ 8pe0 akT je -je = Í ˙ 2 sinhÍ ˙ expÍ ˙ (7) N0 În--m ˚ je Î88pe0 akT ˚ Î pe0 akT ˚ where m± are the positive and negative small ion mobilities, n± their number concentrations, T the temperature, e the modulus of the electronic charge, k Boltzmann’s constant and e0 the permittivity of free space. The mean charge J (Gunn, 1955) is given by 4peakT Èn++ m ˘ J = 0 lnÍ ˙ (8) 2 m e În--˚ Charge distribution on water droplets radius (3.32 ± 0.65)µm, ion ratio 0.82 0.0800 0.0600 N ( j 0.0400 ) / Z 0.0200 0.0000 -40 -35 -30 -25 -20 -15 -10 -5 0 5 10 15 number of particle charges,j experiment MBD theory Figure 7. Charge distribution on water droplets in the presence of ion asymmetry using the Modified Boltzmann Distribution (MBD). (Experimental data from Gunn and Woessner, 1956) 5. OBSERVED TROPOSPHERIC ELECTRICAL PROPERTIES UNDER NON- THUNDERSTORM CONDITIONS Many electrical soundings of the atmosphere have been made during disturbed (thunderstorm) conditions, but few such measurements have been made under more quiescent atmospheric conditions. There are difficulties with in-cloud measurements, as a balloon or aircraft platform will be required: this may itself introduce difficulties with sampling, particularly under small electric fields. Regions of aerosol particles, some of which may acquire appreciable charges, are the principal perturbation to atmospheric electric fields under fair weather conditions. In general the upper and lower surfaces of a horizontal region of particles or droplets will charge from ions flowing vertically as a result of the fair weather conduction current, and the region within the layer will be of low conductivity compared with that of the surrounding air. Figure 8 shows electric field profiles observed in non-thunderstorm clouds, which are typically two orders of magnitude smaller than equivalent profiles determined in thunderstorms. Figure 8. Typical electric field profiles found in non-thunderstorm clouds under (a) liquid water and (b) supercooled conditions (from MacGorman and Rust, 1998). Optical link V to f PLL electrometer SONDE electrodes radio link P,T,U RECEIVER PLL f to V Logger Figure 9. Summary schematic of an atmospheric electric field sensor using vertically-spaced spherical electrodes, flown under a conventional radiosonde balloon. The electrometer circuit and batteries are mounted within the lower electrode, with the data conveyed via a voltage-to-frequency converter and an optical link operating at 100kHz. A phase-locked loop (PLL), recovers the data from an optical receiver, and the signal is injected into a standard meteorological RS80 radiosonde measuring pressure P, temperature T and relative humidity U. The uhf receiver recovers the 100kHz signal, which a further PLL converts back to a voltage. The voltage is logged at 20Hz by a computer and analogue to digital converter (from Harrison, 2001). A modern sensor suitable for use in low fields using a standard meteorological radiosonde has recently been described (Harrison, 2001), using the displacement current to detect charged aerosol particles from the changes caused in electric field. Figure 9 shows a schematic of the system, which is disposable. Regions of space charge of ~10nC.m-3 were reported in shallow layers, with electrical structures suggesting electrification by the conduction current. 6. DIRECT INFLUENCES OF ELECTRIFICATION AND RADIOACTIVITY ON CLOUDS In considering how natural atmospheric ionisation might influence cloud physics or cloud formation, Harrison (2000) identified two possible routes: (1) direct processes, such as production of new aerosol (e.g. sulphate) by gas-to-particle conversion (GPC) or homogeneous nucleation (2) indirect processes, such as the modification of existing heterogeneous nucleation processes by affecting the condensation nuclei (CN) or ice nuclei (IN). 6.1 Direct processes 6.1.1 CN production In the presence of high levels of radioactivity, the radiolytic formation of particles has been shown to occur. Bricard et al (1968) found the CN concentration could be made to cycle by the regular addition of Thoron (a short-lived source of a-particles), figure 10. Vohra et al (1984) observed particle formation in artificial air in the presence of trace concentrations of sulphur dioxide, ozone and ethene, with naturally-occurring radon concentration levels, suggesting that ultrafine aerosol production could occur in atmospheric air under natural conditions. Recent theoretical work by Yu and Turco (2001), further strengthens the expectation that radiolytic particle production will be found in the atmosphere, and that the conventional ion-aerosol balance equations are incomplete. Figure 10. Particles formed in filtered Parisian air, with the addition of regular cycles of Thoron (from Bricard et al, 1968). 6.2 Indirect processes 6.2.1 Scavenging of charged particles The removal of aerosol particles by cloud droplets, scavenging, is known to be influenced by many factors, including electrical forces (Pruppacher and Klett, 1997). Figure 11 shows the partitioning of the water drop charge Qd in response to the image charge induced by the charged aerosol particle (carrying a charge Qa) brought close to the drop. Charge conservation requires that Qd = I + D. In magnitude, I = -(A/s)Qa, where s is the separation distance between the aerosol and drop centres. The image charge is located at a distance c from the centre of the drop (Jackson, 1975). Summing the Coulomb and image forces, the net electrical force acting between the particles' centres is 1 È QI QD˘ F = aa+ (9) e pe Í 22˙ 4 0 Î()bc- b ˚ where a positive Fe is repulsive. CONDUCTING waterdrop charge Qd=Je drop radius A WATERDROP drop charge D c image charge I Charged aerosol trajectory s aerosol charge Qa = j e RADIOACTIVE radioactive decay rate h CHARGED ion pair yield per decay I AEROSOL residual aerosol charge m aerosol radius a Figure 11. Schematic of the (radioactively) charged aerosol and the image charge I construction within a water drop of radius A. The aerosol and falling drop carry charges Qa and Qd respectively, with Q d = I + D , where D is the non-image charge considered at the centre of the drop (from Tripathi and Harrison, 2001). 6.2.2 Electrofreezing Tinsley et al (2000) have shown that the electrical image force is very significant in aerosol- droplet collisions as, unlike the Coulomb force, it is always attractive between the charged aerosol and water droplet at small separations. This process, electroscavenging, is a subset of many processes described more generally as electrofreezing, in which electrical fields or charges influence the freezing of supercooled droplets. Tinsley and Dean (1991) argued that modification of the electrical properties of aerosols might change their efficacy of aerosol as contact ice nuclei, ultimately leading to storm intensification by triggering latent heat release. Direct ionisation has, however, recently been shown not to lead to freezing of supercooled water (Seeley et al, 2001). There is currently no definitive evidence that charging influences ice nuclei efficiency or that contact nucleation of ice is the dominant freezing mechanism. 7. DISCUSSION Ionisation in the atmosphere is ubiquitous and part of the atmospheric electrical system, which transports ions in low electrical field regions of the atmosphere, such as clear air and non- electrified clouds. There appear at least two ionisation-related processes of relevance to cloud formation: (1) aerosol electrification (2) ultrafine aerosol production 7.1 Aerosol electrification Stratified regions of aerosol will charge in the atmosphere under quiescent conditions as a result of ion transport by the conduction current. Although the charges carried are unlikely to be sufficiently large to initiate bulk discharge processes such as lightning, the distribution of charges expected on aerosols under natural ion asymmetry may yield a small fraction of particles with significant charge levels. Such highly-charged aerosol would normally be rapidly neutralised by atmospheric ionisation, but in ion-depleted regions arising from large aerosol concentrations, moderately-high charge levels might persist. Scavenging processes are influenced by aerosol charge; the water drop charge has negligible effect by comparison. Since heterogeneous ice nucleation requires the collection of suitable aerosol able to operate as ice nuclei, it is therefore conceivable that aerosol charging could influence ice formation. If charge were itself shown to be a enhancing effect for ice nuclei, then the synergy between the effects of increased collection and efficient nuclei could be potent. 7.2 Ultrafine particle production Radiolytic particle production has been observed in laboratory air at atmospheric levels of ionisation, but the particles formed are small and not, at their formation, able to act as cloud condensation nuclei. The recent theoretical work of Yu and Turco (2001) is, however, compelling, in that it shows that in aerosol-deficient regions, such as marine stratus cloud, cosmic ray ionisation could provide an appreciable source of particles. Further microphysical modelling is required to show that sufficient ultrafine particles can survive to become cloud condensation nuclei before the effect on suitable clouds can be assessed. 8. CONCLUSIONS The physical processes, if any, leading to the cosmic ray-low cloud correlation observed by Marsh and Svensmark (2000) remain to be established in the atmosphere. As discussed above, there are atmospheric electrical mechanisms relating ionisation to cloud which remain relatively unexplored in atmospheric physics, and in suitable cloud, could conceivably offer physical explanations for the observed correlation. However without numerical and theoretical estimates of their significance, it is currently impossible to regard ionisation effects as irrelevant to cloud processes. REFERENCES Bricard F., Billard F. and Madelaine G., (1968), Formation and evolution of nuclei of condensation that appear in air initially free of aerosols, J. Geophys. Res., 54, 39-52 Chalmers J.A. (1967), Atmospheric Electricity, 2nd edition, Pergamon Press, Oxford Clement C.F. and Harrison R.G. (1992), The charging of radioactive aerosols J. Aerosol Sci. 23, 5, 481-504 Gunn R. (1954), Diffusion charging of atmospheric droplets by ions, and the resulting combination coefficients, J. Meteorol., 11, 339 Gunn R. (1955), The statistical electrification of aerosols by ionic diffusion, J. Coll. Sci., 10, 107- 119 Gunn R. and Woessner R.H. (1956), Measurements of the systematic electrification of aerosols, J. Coll. Sci., 11, 254-259 Harrison R.G. (1997), Climate change and the global atmospheric electrical system Atmos. Environ. 31, 20, 3483-3484 Harrison R.G. (2000), Cloud formation and the possible significance of charge for atmospheric condensation and ice nuclei Space Science Reviews, 94, 381-396 Harrison R.G. (2001), A balloon-carried electrometer for high-resolution atmospheric electric field measurements in clouds Rev Sci Inst, 72, 6, 2738-2741 Harrison R.G. and Lodge B.N. (1998), A calorimeter to detect freezing in supercooled water droplets Rev Sci Inst, 69, 11, 4004-4005 Jackson J.D. (1975), Classical Electrodynamics Wiley Marsh N.D. and Svensmark H. (2000), Low cloud properties influenced by cosmic rays, Phys. Rev. Lett., 85, 23, 5004-5007 MacGorman D.R. and Rust W.D. (1998), The electrical nature of storms, OUP Pruppacher, H.R. and Klett, J.D., (1997). Microphysics of clouds and precipitation, 2nd edition, Kluwer Rogers and Yau (1989), A short course in Cloud Physics, Pergamon Press Seeley L.H., Seidler G.T. and Dash J.G., (2001), Laboratory investigation of possible ice nucleation by ionizing radiation in pure water at tropospheric temperatures, J. Geophys. Res., 106 ,D3, 3033-3036,2001 Tinsley B.A. and Dean G.W (1991) Apparent tropospheric response to MeV-GeV particle flux variations: a connection via electrofreeezing of supercooled water in high-level clouds? J Geophys Res 96, pp22283-22296 Tinsley, B.A., Rohrbaugh R.P., Hei M., and Beard K.V., (2000), Effects of image charges on the scavenging of aerosol particles by cloud droplets, and on droplet charging and possible ice nucleation processes, J. Atmos Sci., 57, 2118-2134. Tripathi S.N. and Harrison R.G. (2001), Scavenging of electrified radioactive aerosol, Atmos Environ, (in press) Vohra K.G., Subba Ramu M.C.and Muraleedharan T.S. (1984), An experimental study of the rôle of radon and its daughters in the conversion of sulphur dioxide into particles in the atmosphere, Atmos. Env., 18, 8, 1653-1656 Volland H. (1984), Atmospheric electrodynamics, Springer-Verlag, Berlin Wilson C.T.R. (1929), Some thundercloud problems, J. Franklin Institute, 208, 1-12 Yu F. and Turco R.P. (2001) From molecular clusters to nanoparticles: the rôle of ambient ionisation in tropospheric aerosol formation J. Geophys. Res. 106, D5, 4797-4814 TROPOSPHERIC ION MEASUREMENTS K L. Aplin* and R. G. Harrison Department of Meteorology, The University of Reading, PO Box 243, Earley Gate, Reading, RG6 6BB UK Abstract To investigate ion-induced nucleation in the atmosphere experimentally, it is necessary to select only the circumstances when ion-induced effects are likely to dominate, and identify and exclude times when other meteorological factors are influencing ion concentrations. To do this, reliable ion, aerosol and meteorological measurements at the same site are required, with analysis of different weather conditions. A methodology for identifying the effects of meteorological conditions on the local atmospheric electrical environment is discussed, based on ion and meteorological measurements made at Reading in the spring and summer of 2000. It is found that even under virtually identical synoptic meteorological conditions, there is significant variability in the ion concentration. 1. INTRODUCTION Atmospheric small ions are produced near the Earth's surface by natural radioactivity and cosmic rays; the electrical conductivity of the air s is proportional to the total ion concentration n. In equilibrium, the concentration of small ions is modulated by the ion production rate and the atmospheric aerosol concentration (e.g. MacGorman and Rust, 1998). Meteorological factors also have an important effect on s, for example, suspended aerosol particles are wind-borne, and may be advected to locally reduce the ion concentration by attachment. The links between meteorology and atmospheric electricity are tacitly acknowledged, but surprisingly little effort has been made to quantify the relationship between them. Chalmers' (1967) discussion was typical of much of the atmospheric electricity literature, clearly recognising meteorological effects on conductivity, but giving little explanation of the causal links. Atmospheric water drops carry an electric charge, and therefore conditions such as fog and haze can cause perturbations to atmospheric electrical measurements, both directly (electrically) and by condensation onto the insulators essential in measuring apparatus. Periods of stable atmospheric electrical conditions, however, are necessary for consistent measurements. In the classical paradigm of “fair-weather” atmospheric electricity, Ohm’s law relates the air conductivity s, vertical charge flux density J and vertical potential gradient E, as J = sE (1) (e.g. MacGorman and Rust, 1998). Conditions under which fair-weather properties can be expected were only summarised relatively recently by Reiter (1992), who excluded periods when hydrometeors were present at the surface. High cloud and fair-weather cumulus were permitted with this classification, until the cumulus started to become grey at the base indicating that it was beginning to charge. However, Barlow and Harrison (1999) showed experimentally that non- electrified clouds perturb the surface atmospheric electric field by thermal influences on the turbulent transport of charged particles and ions. * Now at Rutherford Appleton Laboratory, Space Science and Technology Department, Chilton Didcot, Berks OX11 OQX, UK. The need to investigate meteorological effects on atmospheric ion variability has recently become highly relevant for climate studies, following the published correlation between cosmic ray ionisation and clouds (Svensmark and Friis-Christensen, 1997). A theoretical mechanism linking ions and clouds has been described, involving the nucleation of atmospheric ultrafine aerosol onto atmospheric ions (Yu and Turco, 2001). Explicit observation of this effect in the atmosphere requires the distinction of ion-mediated nucleation from other factors affecting ions, and knowledge of favourable atmospheric conditions for ion-mediated processes. To do this, the effects of particular meteorological conditions on the atmospheric conductivity must be classified. 2. METEOROLOGY AND IONS 2.1 Expected effects of meteorological conditions on the ion concentration Some aspects of the diurnal variation of conductivity (s) (which is directly proportional to the ion concentration) under different meteorological conditions can be inferred. For example, the nocturnal inversion traps radioactive gases near the surface where they cause increased local ionisation and an associated higher conductivity. When the sun rises turbulence sharply increases, mixing the air and dispersing radioactive gases trapped in the surface layer. This “sunrise effect” (Chalmers, 1967) should be most pronounced on clear days, as there is a greater difference between the daytime and nocturnal atmospheric stability. Buoyant convection and the presence of clouds damping turbulence are already thought to influence the atmospheric potential gradient (Barlow and Harrison, 1999), well after sunrise. Smaller “sunrise effects”, showing similar behaviour, may occur throughout the day if the sun is in and out of cloud. Extending similar reasoning, s would be expected to vary more over a clear day than a completely cloudy day, as the former case has a greater diurnal variation in the solar radiation and associated convection. Convective turbulent mixing at the surface reduce ion concentrations to a minimum during the warmest part of the day, suggesting that on cloud-free days there should be an inverse relationship between s and solar radiation. On cloudy days mechanical generation of atmospheric turbulence dominates, with less direct dependence on the solar radiation. 2.2 Ion measurements at Reading Negative conductivity was measured using a self-calibrating instrument (Aplin and Harrison, 2001) at The University of Reading Meteorology Field Site from April-July 2000. Automatic meteorological measurements are made at this site at 1 Hz, and manual weather observations are made daily. Three days are selected for detailed study here: 13th, 17th and 18th June 2000, (year days 165, 169 and 170), for which synoptic charts are shown in Figure 1. These three days would all be traditionally classified as having fair-weather atmospheric electrical conditions. Days 169 and 170 were both characterised by high pressure and weak southerly flow. Day 170 was completely cloud-free and the hottest day of the year, with a maximum of 29.7 ¼C, whereas Day 165 was cooler, with intermittent cloud and westerly flow. a) b) c) Figure 1: Synoptic charts of the three selected days a) 13th June (day 165) b) 17th June (day 169) c) 18th June (day 170) (data from www.wetterzentrale.de). The conductivity was sampled at nominally two-minute intervals and processed as described in Aplin (2000); hourly averages are discussed here and are shown in Figure 2 below. Two cases have been selected for detailed analysis. The cloud-free and hot day 170 is compared with day 165, which had stratocumulus cloud with sunny intervals until 1600 followed by sunshine until 1930. The effect of aerosol on conductivity during cloud-free periods on the consecutive days 169 and 170 is also investigated. 40 165 169 170 30 20 10 negative conductivity (fS/m) 0 45678910111213141516171819202122 hour of day BST Figure 2: Hourly averaged diurnal variation of negative conductivity for 13th, 17th and 18th June 2000 2.3 Conductivity on a cloudy day (day 165) and cloud-free day (day 170) A sunrise effect is clearly apparent on day 170 with a peak at 0400 followed by a decrease in s as the solar radiation increases. The minimum in s occurs between 1600-1700, which coincides with the maximum temperature (1550). Over the whole day, s is negatively correlated to > 95% with the global solar radiation Sg. On day 165, the mean conductivity † (Chilbolton, 48 km SW of Reading) http://www.met.rdg.ac.uk/radar/ some cloud present at Reading from the radiometer measurements, but it is likely to have been thin or high cloud with little impact on the surface conditions. Despite the very similar weather on days 169 and 170 (see Table 1), the conductivities were different and did not show the same variation, with a correlation coefficient r = Ð0.26 between them. This suggests that other processes cause the variability, such as changes in the aerosol concentration. s could also be sensitive to small changes in meteorological variables such as the wind speed. PM10 mass concentrations M are measured at a site 2 km NW of the University field site‡, and are therefore used to infer bulk properties of the air mass. M was almost twice as high on day 169 as day 170, probably because of the increased traffic on a Saturday. The correlation coefficient between M and s was negative on day 169 and positive to > 98% on day 170. Ion-aerosol theory (e.g. Clement and Harrison, 1992) predicts a negative correlation between aerosol and s. Although the theory links aerosol number concentration and s, these observations show the expected general relationship between mass concentration and s. The lower s may therefore have been caused by higher local wind speeds associated with high aerosol concentrations. Table 1: Average conductivity, meteorological conditions, aerosol mass concentration and correlation (r) between M and s for 0530-1830 on 17th and 18th June 2000 -1 Day ‡ http://www.aeat.co.uk/netcen/airqual/ The mass concentration M (mass per unit volume) and the number concentration Z of a monodisperse aerosol population are related by Ê 4 ˆ MZ= Á pr r3 ˜ (4) Ë 3 ¯ where r is the particle density. This equation was used to estimate Z from M assuming the average radius of the population (2 mm) and that the particles are ammonium sulphate, with r = 1.77 gcm-3 (Khlystov et al, 2001). The coefficients a and q were assumed to be 1.6 x 10-6 cm3s-1 and 10 cm- 3s-1 respectively. b was calculated from Gunn’s (1954) expression, and is a function of the mean radius, temperature and mean charge j (assumed to be Ð1e). The aerosol was split into three size modes, nucleation with mean radius rn = 0.25 mm, accumulation with ra = 0.88 mm and coarse with rc = 5.6 mm (Seinfeld and Pandis, 1998). The average sensitivity of the conductivity to a 1% change in the aerosol number concentration in each mode was calculated from equation 5. For the initially assumed size distribution, (shown in Table 2) with 80% of the particles in the fine mode and only 5% in the coarsest size range, the conductivity is most sensitive to changes in the coarse mode. This is not what would intuitively be expected, because there are fewest particles in the coarse mode, and from the ion balance equation (2), the number of particles controls the conductivity, by attachment. Another factor affecting the availability of particles to attach is their surface area. This was tested by fixing the surface area-Z product, by increasing the number of particles in the nucleation and accumulation modes. The nucleation mode was most sensitive to changes in Z; therefore for typical aerosol size distributions, the surface area of the coarse aerosol dominates over the number. Table 2 Characteristics of the assumed aerosol distribution, and calculated sensitivities to a 1% change in aerosol number concentration. Mode Nucleation Accumulation Coarse Assumed fraction of particles 0.8 0.15 0.05 Mean radius (mm) 0.25 0.88 5.6 Sensitivity to 1% change in Z 0.13% 0.11% 0.22% Sensitivity to 1% change in Z 0.2% 0.0% 0.0% (surface area x Z fixed) One finding is that conductivity is relatively insensitive to changes in the aerosol population, which may be significant because of the discussion whether conductivity can be used as an indicator of urban aerosol pollution (see e.g. Aplin, 2000). In this case, changes in the aerosol concentration are not significant enough to account for the observed conductivity variation on day 169, therefore other sources of variability in the conductivity dominate. 3. CONCLUSIONS The conductivity of air in the surface layer is relatively insensitive to the aerosol concentration at the atmospheric concentrations observed; therefore most of its variability is likely to be due to meteorological factors. The different conductivities measured on two days which were a) almost identical synoptically and b) conformed to “fair weather” in the strictest classical sense, imply that the ion concentration is sensitive to small changes in meteorological variables. The traditional “fair-weather” classification, which aims to ensure consistent and comparable measurements, therefore appears inadequate. Furthermore, it does not allow for electro-meteorological interactions and excludes other sources of atmospheric electrical variability. The majority of atmospheric electricity and aerosol measurements made at the surface in England probably occur in non fair-weather conditions, for which there is little understanding of the variability in anything other than general terms. There is evidence that cloud affects surface atmospheric electrical conditions by modulating atmospheric stability. We propose an extension to the existing classification: (1) fair-weather: as Reiter’s (1992) definition, but cloud-free, with variability caused almost entirely by micrometeorological factors (2) semi-fair-weather: presenting similar atmospheric electrical conditions to (1), but identified primarily by meteorological stability criteria with no local charge generation, rather than solely the absence of electrified clouds and (3) non-fair-weather. Extension of the classical fair-weather paradigm to include data obtained on cloudy days, and more detailed investigation of micrometeorological effects on conductivity is important if the ion-induced effects hypothesised to contribute to aerosol production are to be unambiguously identified. ACKNOWLEDGEMENTS The experimental work was supported by the UK Natural Environment Research Council. KLA’s attendance at the IACI meeting was funded by the Environmental Sciences Department, University of Hertfordshire. REFERENCES Aplin K.L. (2000), Instrumentation for atmospheric ion measurements, PhD. Thesis, The University of Reading, UK Aplin K.L. and Harrison R.G. (2001), Rev. Sci. Instrum., 72, 8, in press Barlow J.F and Harrison R.G. (1999), In Christian H.J. (ed) Proceedings 11th International Conference on Atmospheric Electricity, Guntersville, Alabama 7th-11th June 1999 NASA/CP-1999-209261, 575-578 Chalmers J.A. (1967), Atmospheric Electricity, 2nd edition, Pergamon Press, Oxford Clement C.F. and Harrison R.G. (1992), J. Aerosol Sci. 23, 5, 481-504 Dhanorkar S. and Kamra A.K. (1997), J. Geophys. Res. 102, D25, 30147-30159 Dolezalek H. (ed.) (1974), Electrical processes in atmospheres, Springer Verlag, Darmstadt Gunn R. (1954), J. Meteorol., 11, 339 Khlystov A., Kos G. P. A., ten Brink H.M., Mirme A., Tuch T., Roth C. and Kreyling W.G. (2001), Atm. Env., 35, 11, 2045-2051 MacGorman D.R. and Rust W.D. (1998), The electrical nature of storms, OUP Reiter R. (1992), Phenomena in atmospheric and environmental electricity, Elsevier, Amsterdam Seinfeld J.H. and Pandis S.N. (1998), Atmospheric chemistry and physics, Wiley, New York Svensmark H. and Friis-Christensen E. (1997), J. Atmos. Solar-Terrestrial Phys, 59, 1225-1232 Yu F. and Turco R.P. (2001). J. Geophys. Res. 106, D5, 4797-4814. ATMOSPHERIC AEROSOLS: FORMATION AND GROWTH Markku Kulmala University of Helsinki, Department of Physical Sciences, Division of Atmospheric Sciences, P.O. Pox 64, FIN-00014, University of Helsinki, Finland 1. INTRODUCTION It is widely recognised that the increasing atmospheric concentrations of greenhouse gases such as carbon dioxide and methane can potentially drive a significant warming process of the earth’s climate. However, a topic of more recent attention is the possibility that increased atmospheric concentrations of aerosol particles might drive a significant radiative forcing process of the planet (see, for example, Charlson et al., 1992; and Charlson and Wigley, 1994). The increased aerosol concentrations are largely due to secondary particle production i.e. homogeneous nucleation from vapour precursors. The secondary aerosols have both natural and anthropogenic origin. Aerosol particles influence the climate by two distinct mechanisms: the direct reflection of solar radiation by aerosol particles, and the indirect increase in cloud reflectivity caused by enhanced number of cloud condensation nuclei. IPCC (1996) has reported that uncertainties in the estimation of direct and indirect aerosol effects on global climate are big (see Fig. 1.). These uncertainties arise largely from the limited information on the spatial and temporal distribution of aerosols and clouds. However, recently some progress has been made in evaluating the radiative effects of various aerosol components such as sulfate, organics, black carbon, sea-salt, and crustal species (Chuang et al., 1997; Haywood and Ramaswamy, 1998; Kaufman and Fraser, 1997; Winter and Chylek, 1997; Sokolik and Toon, 1996). Despite these efforts, substantial uncertainties still remain in quantifying the contribution from each source, particularly, for biogenic and natural emissions, including organic vapours. Without understanding the contribution of natural emissions of aerosols and particles to radiative forcing, we can never hope to accurately predict or understand the true effect of anthropogenic emissions. Among the key questions in reducing error bars are how aerosol particles are formed, how they will grow from clusters of a few molecules to CCN sizes (>100 nm) and how they will form cloud droplets. Once formed, clouds have a very extensive influence on the Earth's radiation budget through their albedo and greenhouse effects. With global warming, future cloud properties are likely to change due to the warmer and moister conditions, and possibly due to increased aerosol particle emissions from both primary (e.g. wind generated sea-spray) and secondary aerosols (from biogenically and anthropogenically influenced gas-to-particle conversion processes). Clouds are however rather crudely presented in global and regional climate models (GCM, RCM). Processes, such as nucleation, droplet activation during condensation, diffusive growth, droplet evaporation, droplet coalescense and conversion to raindrops, are very crudely taken into account in present-day atmospheric large-scale models. For example, we have recently shown the importance of aerosol formation and growth processes to CCN concentrations (Kulmala et al., 2000) as well as the effect of nitric acid and other semivolatile gases in influencing cloud formation processes, in particular, in enhancing the cloud droplet population, thereby increasing cloud reflectance (Kulmala et al., 1993; Laaksonen et al., 1997). The importance of including multi-component aerosol populations, and the dynamic feedback in the cloud forming processes, along with the importance of coupling chemical and physical processes in predicting cloud droplet populations have been illustrated by O’Dowd et al., (1999a; 1999b). Particle formation and growth in the atmosphere have recently received growing experimental and theoretical interest. Therefore, instrumental techniques for measuring concentrations of freshly formed particle have been developed, and particles with diameter of about 3 nm can be detected. These small particles have been found in large variety of environments: in the free troposphere (Clarke, 1992; Schröder and Ström, 1997; Raes et al., 1997), in the marine boundary layer (Covert et al., 1992; Hoppel et al., 1994; O’Dowd et al., 1998), in the vicinity of evaporating clouds (Hegg et al., 1991), in Arctic and Antarctic areas (Wiedensohler et al., 1996; Pirjola et al., 1998; O’Dowd et al., 1997), in urban areas and in stack plumes (Kerminen and Wexler, 1994; Kerminen and Wexler, 1996; Väkevä et al., 2000). Starting during the mid-nineties, aerosol formation and growth events have been observed also in forested areas e.g. over boreal forest in Finland (Mäkelä et al., 1997, 2000; Kulmala et al., 1998), and in other type of forests in Portugal (Kavouras et al., 1998), Greece (Kavouras et al., 1999), Canada (Leaitch et al., 1999), and in USA (Marti et al., 1997). In all these cases particle formation and growth events took place in remote forested areas, where the release of highly reactive volatile organic carbons (VOCs) from trees followed by a rapid oxidation to low volatile products, has to be considered as a potential source for nucleating vapours. 3 Tropospheric ) Halocarbons 2 aerosols - direct effect 2 N 2 O C H 4 Tropospheric Fossil 1 C O 2 Ozone fuel Solar soot Biomass Sulphate burning 0 Stratospheric -1 Ozone Tropospheric aerosols -indirect effect -2 Global mean radiative forcing (W /m Confidence level Very Very High L o w Low Low Very Very Low Low Low Low Figure 1: Estimates of globally and annually averaged anthropogenic radiative forcing (in Wm-2) due to the changes in concentrations of greenhouse gases and aerosols from pre-industrial times to present day and to natural changes in solar output from 1850 to present day (IPCC, 1996). Atmospheric aerosol particles in urban areas, on the other hand, cause the loss of visibility (e.g. Finlayson-Pitts and Pitts, 2000) and health effects (Dockery and Pope, 1994). Heavily industrialized areas suffer from pollution fogs (smogs) that are often related to coal burning and nowadays also to traffic. The most well-known example of such smogs is the London ”pea- souper” smog, which occurred every once in a while until the 50’s, when coal burning was forbidden. Besides visibility degradation, the London smog episodes caused serious health effects and ”excess deaths”. One significant part of health problems related to atmospheric aerosols and fog droplets, since particles having diameters less than 10 mm can penetrate deep into the respiratory system (Dockery and Pope, 1994). Recently, the effect of ultra-fine particles have been discussed and their local variations have been investigated (e.g. Buzorius et al., 1999). 2. AEROSOL DYNAMICS During the processes of formation and growth of atmospheric aerosols the aerosol dynamics, atmospheric chemistry and meteorology form a coupled system. The importance of atmospheric chemistry (e.g. Pirjola and Kulmala, 1998; Pirjola 1999) as well as meteorological conditions (Nilsson and Kulmala, 1998; Nilsson et al., 2000; Väkevä et al., 2000) on particle formation and growth have been demonstrated under tropospheric conditions. Although ternary nucleation of water-ammonia-sulphuric acid vapours (Korhonen et al., 1999) has shown to be able to explain atmospheric nucleation Ð i.e. formation of ~1 nm particles - in many cases (Kulmala et al., 2000), the exact routes for formation of 3 nm particles are still unclear, because besides nucleation, also the growth from 1 nm size to 3 nm size is needed. In order to be able to understand the formation and growth processes of atmospheric aerosols and cloud droplets their thermodynamic properties should be known. For example, in the condensation process, the driving force is the vapour pressure difference between gas phase and surface. However, in the atmosphere where there are multicomponent, multiphase mixtures, their thermodynamic state and phase diagrams are typically very complex. It is very important to obtain thermodynamically consistent vapour pressures, chemical activities, surface tensions and densities for organic and inorganic compounds and their water solutions (for the importance see e.g. Korhonen et al., 1999) as a function of temperature and composition. In future, development of nucleation theories, modelling and nucleation rate parameterizations are needed. So far, conclusions on whether or not certain substances cause nucleation in the atmosphere conditions are usually based on predictions given by the classical nucleation theory (CNT). CNT treats the nucleating molecular clusters as macroscopic droplets which is a questionable approach since the nucleating clusters often contain less than fifty molecules. Nucleation of various vapors using molecular dynamics (MD) and Monte Carlo (MC) simulation techniques is needed to investigate. So far, some investigations were carried out using ab initio calculations on small sulfuric acid-water clusters (Arstila et al 1998), classical MD (Laasonen et al, 2000) and MC (Vehkamäki and Ford, 1999) simulations of argon nucleation, as well as DFT calculations of nucleation in binary systems imitating water and different organic molecules (Laaksonen et al., 1995, Napari and Laaksonen 2000). Also, a new nucleation mechanism based on stable dimers (Lushnikov and Kulmala, 1998) has been proposed. In contrast to laboratory conditions, the formation of aerosol in the atmosphere can be kinetically limited by some of the intermediate steps of its formation processes. The equilibrium state is thus not necessarily the aerosol itself but can be, for example, thermodynamically stable clusters (TSC), as we have recently shown (Kulmala et al., 2000). Although there is strong indication that the water-sulphuric acid-ammonia nucleation mechanism (Korhonen et al., 1999) explains the formation of new atmospheric aerosols (diameter < 3 nm) in many circumstances, the condensation of these vapors does not explain the observed growth rates of the particles (Kulmala et al., 2000), and in atmospheric conditions nucleation and growth are decoupled (Kulmala et al., 2000). The other possible relevant nucleation mechanism is ion-induced nucleation. Aerosol dynamic modelling (nucleation, condensation, coagulation, deposition) with gas phase chemistry to obtain the atmospheric significance of nucleation and condensation of different vapours have been and will be performed. The aerosol dynamics and atmospheric chemistry model used in the present research is based on the model recently developed by our research group (Pirjola and Kulmala, 1998; Pirjola 1999). In these models aerosol formation and growth including aerosol dynamics to evaluate sink terms for condensable molecules and gas phase chemistry to include source terms for these molecules will be used. Process models will be coupled with dispersion models. In the chemistry part of the model the chemistry of O3, NOx, VOC and other relevant species will be related to aerosol formation. The effects of meteorological dynamics on aerosol processes will be studied by applying the aerosol dynamic models in a Lagrangian approach including wave motions and atmospheric mixing. The results shows that ternary water-ammonia-sulphuric acid system is proper candidate for atmospheric aerosol formation. 3. FORMATION AND GROWTH OF ATMOSPHERIC AEROSOLS, FIELD EXPERIMENTS Formation and growth of aerosol particles have been observed and will be observed at atmospheric conditions. Our research group has participated in several field campaigns. These includes continuous measurements performed at our field stations and several international intensive campaigns like Aerosol Characterisation Experiment 1 and 2 in 1994 and 1997 (ACE-1 and ACE- 2 organised by IGAC), International Arctic Ocean Expeditions 1991 and 1996, Biogenic aerosol formation in the boreal forest (BIOFOR, 1997-1999, SMEAR stations, Finland, Hyytiälä), New particle formation and fate in the coastal environment (PARFORCE, Mace Head, Ireland), and ongoing the OSOA (Origin and Formation of Secondary Organic Aerosol) experiment. As an example we consider here BIOFOR results in more detailed. All data measured during the BIOFOR campaigns are available on the Biofor web pages http://mist.helsinki.fi/Biofor/index.html (ask for usercode and password from the corresponding author). In addition to the numerical data there are also a number of plots produced as a result of the analysis of the data. The data are classified into 9 subgroups: 1) aerosol total number concentration and size distribution measurements in the size range 3-800 nm, 2) aerosol chemistry, 3) aerosol and gas fluxes by eddy covariance and gradient methods, 4) measurements of meteorological parameters and gas concentrations at six different levels from the mast, 5) meteorology of boundary layer and trajectories, 6) concentrations and emissions of BVOC (biological volatile organic compounds), 7) ground level concentrations of inorganic gases, 8) measurements of the size distribution of wet (ambient) aerosol from 0.5-32 _m at 18 m height, and 9) solar radiation measurements. The detailed descriptions of the instruments used are given on the web pages. When the particle formation event occurs, the mode of the fresh particles appears into the measurement range. In Figure 2. aerosol number size distributions measured using Differential Mobility Particle Sizer (DMPS) during a typical nucleation event day are shown. The nucleation mode practically dominates the spectrum with its high number concentration during the nucleation burst. For this event, particle growth from nucleation mode up to accumulation mode is clearly observable. The growth is frequently seen to continue during the following days up to accumulation mode (see also Kulmala et al., 2001). 105 104 103 00 03 06 09 12 15 18 21 00 Time of day (hr) Figure 2: Aerosol size distributions measured by DMPS from 2m height inside the forest (6.4. 1999). 105 TSI3010 67m TSI3025 67m 104 103 00 03 06 09 12 15 18 21 00 Time of day (hr) Figure 3: Aerosol concentrations measured by CPC’s above the forest heights 67m and 18m (6.4. 1999). During the events, aerosol fluxes determined using an eddy covariance technique are observed to be downwards. Also the measurements made by Condensation Particle Counters (CPC) and DMPS at different heights support this finding. From particle flux data, using the eddy covariance method (Buzorius et al., 1998), usually a small overall downward flux is observed. The downward flux clearly increases during nucleation events, with an exception of the cases when the surface wind was from direction of 220-250∞ (direction of the Tampere city and the Hyytiälä institute buildings). Then a strong upward particle flux is observed due to local surface-level pollution. The difficulties in absolute calibration of the DMPS set ups as well as sampling losses in the lines suggested that the gradient of particles will be best determined placing two identical CPC pairs in the mast (18 m and 67 m height). The CPC pairs consisted of the ultrafine CPC (TSI Inc 3025) for determination of the particles larger than 3 nm in diameter and conventional CPC (TSI Inc. 3010) for particles larger than 10 nm in diameter. The difference of the reading of the CPC’s gives an approximate value for the ultrafine mode particle concentrations in the beginning of the burst. The data from the CPC pairs is shown for the event day of 6 April 1999 in Figure 3. The difference between the CPC readings from the two levels shows that the ultrafine particles have higher concentrations in higher level during the nucleation burst. This result will support the particle flux data that illustrate a net loss of particles to the canopy; however, it does not necessarily indicate a particle source at the top of the boundary layer or higher altitudes, even though nucleation is more probable in these regions. 4. CONCLUSIONS According to recent results on atmospheric aerosol formation some preliminary conclusions can be made on atmosphere aerosol formation. (see Kulmala et al., 2001) The most probable formation mechanism is ternary nucleation (water Ð sulphuric acid Ð ammonia) and the growth to observable sizes takes place mainly owing to condensation of organic vapours. Nevertheless, there is no direct proof of this phenomenon because the composition of 1Ð5 nm size particles is very difficult to determine using present state-of-art instrumentation. The other possible nucleation mechanism is ion induced nucleation with sulphuric acid and water vapours. Nucleation takes place typically in very specific weather conditions: e.g. in Hyytiälä in cold air advection in Polar and Arctic air masses, at low cloudiness, and no precipitation. Furthermore, the nucleation was closely connected to the onset of strong turbulence in the morning-noon transition from stable to unstable stratification, which should also correspond to the onset of convection and entrainment from aloft. The emission rates for several gaseous compounds have been determined (Kulmala et al., 2001). Using four independent ways the amount of the condensable vapour needed for observed growth of aerosol particles was estimated to 2-10 x 107 vapour molecules cm-3. The estimations for source rate gives 7.5-11 x 104 cm-3s-1. REFERENCES Arstila H., Laasonen K. and Laaksonen A, (1998) J. Chem. Phys., 108, 1031. Buzorius, G., U. Rannik, J.M. Mäkelä, T. Vesala and M. Kulmala (1998) J. Aerosol. Sci. 29, 157. Buzorius, G., Hämeri, K., Pekkanen, J. and Kulmala M. (1999) Atmospheric Environment 33, 553. Buzorius, G., Ü. Rannik, J. M. Mäkelä, P. Keronen, T. Vesala, M. Kulmala. (2000) J. Geophys. Res. 105, 19905. Charlson, R.J., and Wigley, T.M.L. 1994. Sulphate aerosol and climatic change. Scientific American, 270, 48-57. Charlson, R.J., Schwartz, S.E., Hales, J.M., Cess, R.D., Coakley, J.A., Hansen, J.E. and Hofmann, D.J. 1992. Climate forcing by anthropogenic aerosols. Science, 255, 423-430. Clarke, A.D. 1992. Atmospheric nuclei in the remote free troposphere. J. Atmos. Chem., 14, 479- 488. Covert, D.S., Kapustin, V.N., Quinn, P.K., and Bates, T.S. 1992. New particle formation in the marine boundary layer. J. Geophys. Res., 97, 20581 Ð 20587. Chuang C. C., Penner J. E., Taylor K. E., Grossman A. S. and Walton J. J. (1997). J. Geophys. Res. 102, 3761. Dickerson, R.R., Kondragunta, S., Stenchikov, G., Civerolo, K.L., Doddridge, B.G. and Holben, B.N. (1997) Science 278, 827. Dockery, D.W. and Pope, C.A. (1994) Annu. Rev. Public. Health, 15, 107. Finlayson-Pitts B.J. and Pitts, Jr. J.N. (2000) Chemistry of the upper and lower atmosphere. Academic Press, San Diego. Hämeri, K., Väkevä, M., Hansson, H.-C. and Laaksonen A. (2000) J. Geophys. Res. 105, 22231. Haywood J. M. and Ramaswamy V. (1998) J. Geophys. Res. 103, 6043. Hegg, D.A., Radke, L.F. and Hobbs, P.V. 1991. Measurements of Aitken nuclei and cloud condensation nuclei in the marine atmosphere and their relation to the DMS-cloud-climate hypothesis. J. Geophys. Res., 96, 18727-18733. Hoppel, W.A., Frick, G.M., Fitzgerand, J.W. and Larson, R.E. 1994. Marine boundary layer measurements of new particle formation and the effects nonprecipitating clouds have on aerosol size distribution. J. Geophys. Res., 99, 14443-14459. IPCC (Intergovernmental Panel on Climate Change) 1996. Climate Change 1995 (Edited by Houghton J.T., Meira Filho L.G., Gallander B.A:, Harris N., Kattenberg A. and Maskell K.). Cambridge University Press, Cambridge Kaufman Y. J. and Fraser R. S. (1997) Science 277, 1636. Kavouras, I.G., Mihalopoulos, N. and Stephanou, E.G. 1998. Formation of atmospheric particles from organic acids produced by forests. Nature, 395, 683-686 Korhonen, P., Kulmala, M., Laaksonen, A., Viisanen, Y., McGraw, R. and Seinfeld, J.H. 1999. Ternary nucleation of H2SO4, NH3 and H2O in the atmosphere. J. Geophys. Res., 104, 26,349-26,353. Kulmala, M., A. Laaksonen, P. Korhonen, T. Vesala, T. Ahonen, J.C. Barrett, (1993) J. Geophys. Res. 98, 22949. Kulmala, M., Laaksonen A., Charlson R.J. and Korhonen P. (1997) Nature, 388, 336. Kulmala, M., A. Toivonen, J. M. Mäkelä and A. Laaksonen, 1998. Analysis of the growth of nucleation mode particles observed in Boreal forest. Tellus, 50B, 449-462. Kulmala, M., Pirjola, L. and Mäkelä, J.M. 2000. Stable sulphate clusters as a source of new atmospheric particles. Nature, 404, 66-69. Kulmala, M., K. Hämeri, P.P. Aalto, J.M. Mäkelä, L. Pirjola, E.D. Nilsson, G. Buzorius, U. Rannik, M. dal Maso, W. Seidl, T. Hoffman, R. Janson, H.-C. Hansson, Y. Viisanen, A. Laaksonen, and C.D. O’Dowd. 2001. Overviwe of the internal project on biogenic aerosol formation in the boreal forest (BIOFOR). Tellus 53B, 324-343. Laaksonen, A. Talanquer, V., and Oxtoby, D.W (1995) Annu. Rev. Phys. Chem., 46, 489. Laaksonen, J. Hienola, M. Kulmala, F. Arnold, (1997) Geophys. Res. Lett. 24, 3009. Laaksonen, P. Korhonen, M. Kulmala, R.J. Charlson, (1998) J. Atmos. Sci. 55, 853. Laasonen, K., , S. Wonzack, R. Strey and A. Laaksonen, J. Chem. Phys. (in press) Leaitch, W.R., Bottenheim, J.W., Biesenthal, T.A., Li, S.M., Liu, P.S.K., Asalian, K., Dryfhoutclark, H., Hopper, F. and Brechtel, F. 1999. A case study of gas-to-particle conversion in an eastern Canadian forest. J. Geophys. Res. 104, 8095-8111. Lushnikov, A.A., Kulmala, M. (1998) Phys. Rev. Lett. 81, 5165. Marti, J.J., Weber, R.J., McMyrry, P.H., Eisele, F., Tanner, D., and Jefferson A. 1997. New particle formation at a remote continental site: Assessing the contributions of SO2 and organic precursors. J. Geophys. Res. 102, 6331-6339. Mäkelä, J.M., Aalto, P., Jokinen, V., Pohja, T., Nissinen, A., Palmroth, S., Markkanen, T., Seitsonen, K., Lihavainen, H., and Kulmala, M. 1997. Observations of ultrafine aerosol particle formation and growth in boreal forest, Geophys. Res. Lett., 24, 1219-1222. Mäkelä, J.M., Koponen, I.K., Aalto, P. and Kulmala, M. 2000. One-year data of submicron size modes of tropospheric background aerosol in Southern Finland. J. Aerosol Sci., 31, 595- 611. Napari, A. Laaksonen, L. (2000) Phys. Rev. Lett. 84, 2184. Nilsson, E.D. and Kulmala, M. 1998. The potential for atmospheric mixing processes to enhance the binary nucleation rate. J. Geophys. Res., 103, 1381-1389. Nilsson, E.D., Pirjola, L. and Kulmala, M. 2000. The effect of atmospheric waves on aerosol nucleation and size distribution. J. Geophys. Res., 105, 19,917-19,926. O’Dowd, C.D., Lowe, J.A., Smith, M.H., Davison, B., Hewitt, C.N. and Harrison, R.M. 1997. Biogenic sulphur emissions and inferred non-sea-salt-sulphate cloud condensation nuclei in and around Antarctica, J. Geophys. Res., 102, 12,839-12,854. O’Dowd, C.D., Geever, M., Hill, M.K., Jennings, S.K., and Smith, M.K. 1998. New particle formation and spatial scales in the clean marine coastal environment. Geophys. Res. Lett., 25, 1661-1664 O’Dowd, C. D., J. A. Lowe, M. H. Smith and A. D. Kaye (1999a) Q. J. R.. Meteorol. Soc., 125, 1295. O'Dowd, CD, J.A. Lowe, and MH Smith (1999b) Geophys. Res. Letts, 26, 1311. Seckmeyer, G., Mayer, B., Bernhard, G., Erb, R., Albold, A., Jäger, H. and Stockwell, W.R. (1997) Atmospheric Environment 34, 2971. Pirjola, L. 1999. Effects of the increased UV radiation and biogenic VOC emissions on ultrafine sulphate aerosol formation. J. Aerosol Sci., 30, 355-367 Pirjola, L. and Kulmala, M. 1998. Modelling the formation of H2SO4-H2O particles in rural, urban and marine conditions. Atmos. Res., 46, 321-347. Pirjola, L., Laaksonen, A., Aalto, P., and Kulmala, M. 1998. Sulfate aerosol formation in the Arctic boundary layer. J. Geophys. Res., 103, 8309-8322. Raes, F., Van Dingenen, R., Cuevas, E., Van Velthoven, P.F.J., and Prospero, J.M. 1997. Observations of aerosols in the free troposphere and marine boundary layer of the subtropical Northeast Atlantic: Discussion of processes determining their size distribution, J. Geophys. Res., 102, 21315-21328. Schröder, F. and Ström, J. 1997. Aircraft measurements of submicrometer aerosol particles (> 7 nm) in the midlatitude free troposphere and tropopause region. Atmos. Res., 44, 333-356. Seinfeld, J.H. and Pandis (1998) Atmospheric Chemistry and Physics. (Wiley) Sokolik I. N. and Toon O. B. (1996) Nature 381, 681. Vehkamäki, H., and I. J. Ford, (1999) J. Chem. Phys., 112, 4193. Viisanen, Y. (1991) Ph.D.Thesis, University of Helsinki, Department of Physics. Viisanen, Y., Strey, R., Reiss, H., (1993) J. Chem. Phys., 99, 4680. Viisanen, Y., Strey, R., Reiss, H. (2000) J. Chem. Phys., 112, 8205 Viisanen, Y., Strey, R. (1996) J. Chem. Phys., 105, 8293 Viisanen, Y., Kulmala, M., Laaksonen, A., (1997) J. Chem. Phys., 107, 920 Väkevä, M., Hämeri, K., Puhakka, T., Nilsson, E.D., Hohti, H. and Mäkelä, J.M. 2000. Effects of meteorological processes on aerosol particle size distribution in an urban background area, J. Geophys. Res., 105, 9807-9821. Winter B. and Chylek P. (1997)Tellus 49B, 72. COSMIC RAYS, PARTICLE FORMATION, NATURAL VARIABILITY OF GLOBAL CLOUDINESS, AND CLIMATE IMPLICATIONS Fangqun Yu Atmospheric Sciences Research Center, State University of New York, Albany, New York, USA Abstract Via its role in aerosol formation, cosmic ray may affect the global cloudiness and hence climate. Here we show that an increase in cosmic ray fluxes may lead to an increase in particle production in the lower troposphere but a decrease in particle production in the upper troposphere. In addition to the reported positive correlation between cosmic ray variations and low cloudiness, our analysis of satellite-based cloud cover data reveals that high cloudiness may be anti-correlated with cosmic ray variations if volcano and El Niño impacts are excluded. The observed different correlations between cosmic ray variations and low, middle and high cloud anomalies are consistent with the predicted different sensitivities of particle production to cosmic ray changes at various altitudes. The influence of the solar-modulated cosmic ray fluxes on global cloudiness, if confirmed, may provide the external forcing needed to reconcile the apparent differences between observed surface and troposphere temperature trends. 1. INTRODUCTION Clouds play a key role in the energy budget of Earth’s surface and lower atmosphere, and are probably the largest contributor to the uncertainty concerning the global climate change1. Small modifications of the amount, distribution, or radiative properties of clouds can have significant impacts on the predicted climate2. To detect and attribute anthropogenic influences on climate, it is crucial to quantify the natural fluctuations of cloudiness and the associated radiative forcing. In 1997, Svensmark and Friis-Christensen3 reported a surprising discovery that total cloud cover over midlatitude ocean correlates closely with the galactic cosmic ray (GCR) intensity. The cloud data analyzed include the C2 data sets from the International Satellite Cloud Climatology Project (ISCCP)4. Recently it has become possible to infer global cloud properties at different altitudes from the ISCCP-D2 data, which come from an improvement of procedures leading to the C2 data5. Analyses of the ISCCP-D2 data indicate that a clear correlation can only be seen between GCR fluxes and the global average of low cloud cover6,7. Due to its potential importance and implication, the GCR-cloud-climate hypothesis3 has been under close scrutiny8-10. Two of the main questions and doubts raised against the hypothesis are: (1) no convincing physical mechanism is available to explain the correlation, (2) there is no obvious correlation between solar activity and high cloudiness (where, it is argued, if GCR ionization has any impact on cloud microphysics, it would most likely be found in the upper troposphere where GCR incidence is greatest). Here we first try to address these two raised issues by investigating the role of GCR ionization in particle formation and the potential altitude-dependent influence of GCR variations on particle production and global cloudiness. We then explore the possibility of GCR-induced global cloud changes as an external forcing that may reconcile the apparent differences in global mean temperature trends between ground and atmosphere measurements. Over the last two decades, the temperature records taken at the Earth's surface show rapid warming (globally 0.15 ± 0.05 oC per decade), however the data produced by satellite and balloon studies indicate little if o any warming (globally 0.05 ± 0.10 C per decade) of the low to mid-troposphere - the atmospheric layer extending up to about 8 km from the Earth's surface11,12. Climate models generally predict that this atmospheric layer should warm faster than the surface if increased concentrations of greenhouse gases are causing the warming. Model simulations taking into account the effects of sulfate aerosols, stratospheric ozone depletion, and volcano eruptions were not able to reconcile these inconsistencies13-15. Gaffen et al16 suggested that these inconsistencies may be associated with external forcings of climate system that result in different surface and low tropospheric temperature changes. It is of interest and importance to investigate if the GCR- induced global cloud changes can provide the kind of external forcing needed to reconcile the inconsistencies. 2. GCR-CN-CCN-CLOUD HYPOTHESIS GCR-CN-CCN-Cloud Hypothesis GCR CN CCN Cloud GCR Variations Production rate Abundance Properties Ions Clusters CN CCN Cloud Droplets Ion Condensation Condensation Activation Nucleation Coagulation Coagulation H2SO4 H2SO4, Organics Ions, H2SO4 (?? NH3, Organics) H2O SO2 (?? NH3, HNO3, HCl) DMS Figure 1. Schematic illustrating of GCR-CN-CCN-Cloud Hypothesis that, if confirmed, might explain the correlation between variations of GCR flux and low cloud cover. The possible dominating species involved in the different phases of CN formation and growth processes are also indicated. The organics species may play an important role in growing the CN into the size of CCN. Figure 1 shows the GCR-CN-CCN-Cloud hypothesis that, if confirmed, might offer a physically- based link between GCR fluxes and global low-level cloud properties. Several steps are involved in this hypothesis. First, the modulation of galactic cosmic radiation by the solar cycle will cause a notable variation in aerosol production and condensation nuclei (CN) population in the lower atmosphere. Second, a systematic change in the ultrafine particle production rate will affect the population of cloud condensation nuclei (CCN). Third, a change in CCN abundance will affect the cloud properties. Clouds that form in air containing high CCN concentrations tend to have high droplet concentrations, which leads to both an increase in albedo and an increase in absorption. Increase in the CCN concentration also inhibits rainfall and therefore increases cloud lifetimes (cloud coverage). These effects Ð which are due to more, smaller droplets at fixed liquid water content Ð are particularly significant in marine air, where the CCN concentrations are generally quite low. In our proposed hypothesis, the first key process connecting GCR flux and low cloud cover is that ions generated by GCR ionization play an important role in new particle formation in the lower atmosphere which is the focus of discussion of the next section. 3. COSMIC RAY IONIZATION AND PARTICLE FORMATION Ambient ions are continuously generated by galactic cosmic rays at the rate of ~2 ion-pairs cm-3s-1 at ground level and up to ~20-30 ion-pairs cm-3s-1 in the upper troposphere17,18. Due to enhanced growth and stability of charged clusters (as a consequence of electrostatic interactions), air ions produced from GCR ionization may play an important role in the production of new particles under typical tropospheric conditions19,20. The proposed ion-mediated nucleation (IMN) theory can physically explain the enhanced growth rate (a factor of ~ 10) of sub-nanometer clusters and 2 the square of sulfuric vapor concentration ([H2SO4] ) dependence of nucleation rate as observed by Weber et al21, and seems to account consistently for ultrafine aerosol formation in jet plumes, in clean continental air and in marine boundary layer, as well as for the diurnal variation in the atmospheric mobility spectrum, as demonstrated by Yu and his colleagues 19,20,22-24. It is generally known that sulfuric acid vapor concentration ([H2SO4]), temperature (T), relative humidity (RH), pressure (P), and the surface area of preexisting particles are among the list of parameters controlling the particle formation in the troposphere. The IMN theory adds another important parameter-ion concentration ([ion], or ionization rate Q)-to this list. Here we focus on investigating the influence of GCR variations on particle formation and CN abundance at different altitudes. We employ an advanced particle microphysics (APM) model that simulates a size-resolved multicomponent aerosol system via a unified collisional mechanism involving both neutral and charged particles down to molecular sizes20. The size-resolved ion-ion recombination coefficients, ion-neutral collision kernels, and neutral-neutral interaction coefficients calculated in the model are physically consistent and naturally altitude (temperature, pressure, and relative humidity) dependent20. For the simulations presented below, the ion concentration is initialized as Q /a where a is ion-ion recombination coefficient. The pre-existing particles are initialized as two log-normal modes with total number densities of 19.5/cm3 and 0.6/cm3, median dry diameters of 0.09 mm and 0.3 mm, and standard deviations of 1.6 and 1.5, respectively. This gives an initial 2 3 wet surface area of ~ 4.5 mm /cm at 90% relative humidity, corresponding to a cloud-processed clean air mass where typical significant aerosol nucleation has been observed. Figure 2 shows the total condensation nuclei bigger than 3nm (Nd>3 nm) after three hours of simulations as a function of ionization rates at three different altitudes (0, 5, 8 km). The values of [H2SO4], T, P, and RH for each altitude (as specified in the figure legend) are fixed during the three-hour simulations. The shaded areas in Figure 1 are the ranges of Q values corresponding to low (>680 mb), middle (440-680 mb), and high (<440 mb) cloud regions as defined in ISCCP cloud data according to the cloud top pressures. It is clear from Figure 2 that significant number of ultrafine particles have formed under all the considered conditions. Most of these newly formed particles began as electrically charged clusters that have the advantage of enhanced growth and stability due to electrostatic effects. The neutral sub-critical clusters, on the other hand, grow too slowly to exceed the critical size under the prevailing conditions. The production rate of ultrafine particles is most sensitive to [H2SO4] and [ion] (or ionization rate). [H2SO4] controls the growth rate of ion clusters, while [ion] determines the lifetime of charged clusters and the availability of ions. The neutralization by ion-ion recombination will make the growing charged clusters lose their growth advantage and the resulting neutral clusters may dissociate if smaller than the critical size. At typical [H2SO4] where nucleation has been observed, for very low Q most of the ion clusters have sufficient time to reach the larger stable sizes prior to recombination and the nucleation rate is limited by Q. As Q increases, ion concentration increases but the lifetime of ions decreases and hence the fraction of ions having sufficient time to grow to the larger stable sizes decreases. As a result, the total number of particles nucleated first increases but later on decreases as Q increases. Figure 2 demonstrates that, as Q increases, Nd>3 nm increases rapidly in the low cloud region but decreases in the high cloud region. The Q value at turning point (i.e., dN/dQ=0) is sensitive to [H2SO4] and is most likely located in middle cloud region. 7 3 0 km (T=288 K, P=1013 mb, RH=90%) with [H2 SO 4 ]= 2.0x10 / cm 7 3 5 km (T=256 K, P= 541 mb, RH=77%) with [H2 SO 4 ]= 1.5x10 / cm 7 3 8 km (T=236 K, P= 357 mb, RH=68%) with [H2 SO 4 ]= 1.1x10 / cm 20000 18000 16000 5 km 0 km 14000 ) 3 12000 (#/cm 10000 d>3 nm N 8000 6000 8 km 4000 Low Middle High 2000 051015 20 25 30 Ionization rate Q (ion-pairs/cm3) Figure 2. Simulated concentrations of total condensation nuclei larger than 3nm (Nd>3 nm) after three hour of simulations for various ionization rates (Q) at three altitudes (0, 5, and 8 km). The shaded areas are the ranges of Q corresponding to low (>680 mb), middle (440-680 mb), and high (<440 mb) cloud regions as defined in ISCCP cloud data. Nd>3 nm increases rapidly in the low cloud region but decreases rapidly in the high cloud region as Q increases. During a solar cycle, the values of Q vary by ~20-25% in the upper troposphere and ~5- 10% in the lower troposphere. To study the effect of such systematic change of ionization rates on particle production at different altitudes, we increase the baseline ionization rate at each chosen altitude by 20% and compare the CN abundance after three hours of simulations. The altitude- dependent values of [H2SO4], Q, T, RH, P, and the surface area of preexisting particles are specified and some of them are shown in Figure 3. The baseline values of Q at different altitudes are from observations17,18, and the temperature and pressure are according to the US standard atmosphere. The [H2SO4] and RH are parameterized in a way so that they are constant in the lowest 2 km of atmosphere (2x107/cm3 and 90%, respectively) and gradually decrease with altitude above 2 km. These parameterizations are reasonable and are within the range of the observed values in various field campaigns25,26. Figure 4 shows the total condensation nuclei bigger than 3nm (Nd>3 nm) after three hours of simulations at different altitudes. The black line (with opaque circles) is for the baseline Q values while the green line (with filled circles) is for Q values 20% over the corresponding baseline values. The shaded areas in Figure 4 are low, middle, and high cloud regions as defined in ISCCP cloud data. [H2SO4], Q, T, and RH at each altitude (see Figure 3) are fixed during the three-hour simulations. It is clear from Figure 3 that an increase in GCR ionization rate associated with solar activity leads to an increase in the ultrafine production rate (i.e., dN/dQ>0) in the lower troposphere (as indicated by the red arrows) but a decrease in the ultrafine production rate (i.e., dN/dQ<0) in the upper troposphere (as indicated by the blue arrows). In the middle troposphere, dN/dQ changes sign and the average value of dN/dQ is small compared to that of lower and upper troposphere. It is interesting to note that the optimum particle formation layer is located in the middle troposphere (3-5 km altitude, likely in cloud outflows or top of low clouds), which is consistent with the measurements obtained in recent field campaigns such as ACE-126. 3 T (K) H2SO4 (#/cm ) 190 210 230 250 270 290 5x106 1077 1.5x10 2x107 2.5x107 12 12 11 11 H2SO4 10 10 9 9 8 T 8 Altitude (km) 7 7 6 6 5 RH 5 Altitude (km) 4 4 3 3 2 Surface Area/5 Q 2 1 1 0 0 0.5 0.6 0.7 0.8 0.9 1.0 051015 20 25 30 35 40 2 3 3 Relative Humidity, Surface Area/5 ( µm /cm ) Q (ion-pairs/cm s) Figure 3. The vertical profiles of [H2SO4], baseline Q, T, and RH used in the model to study the effect of a systematic change in ionization rates on particle production at different altitudes. 12 11 10 9 High 8 7 440 mb 6 5 Middle Altitude (km) 4 3 680 mb 2 Low 1 0 0 5000 10000 15000 3 Nd>3 nm (#/cm ) Figure 4. Simulated concentrations of total condensation nuclei larger than 3nm after three hours of simulations at different altitudes. The black line is for the baseline Q values while the green line is for Q values 20% over corresponding baseline values. The arrows indicate the changes in Nd>3 nm as ionization rates increase by 20%. The shaded areas are ISCCP low, middle, and high cloud regions. 4. NATURAL VARIABILITY OF GLOBAL CLOUDINESS It is well known that the abundance of cloud condensation nuclei (CCN) affects cloud properties27- 29. Clouds that form in air containing high CCN concentrations tend to have high droplet concentrations, which lead to an increase in both cloud albedo and absorption. Increases in the CCN concentration also inhibit rainfall and therefore increase cloud lifetimes (cloud coverage). Since the dominating number of CCN is evolved from newly formed ultrafine particles, a systematic change in the ultrafine particle production rate will affect the population of CCN. It is physically plausible that an increase in ultrafine production rate will increase the CCN abundance and cloudiness. During a solar cycle, the values of Q vary by ~20-25% in the upper troposphere and ~5-10% in the lower troposphere. Therefore, based on the influence of GCR ionization change on particle formation rate at different altitudes as shown in Figure 4, we can expect that if GCR variations have any impact on cloudiness, they should correlate positively with low cloud amount and negatively with high cloud amount. For middle clouds, such a correlation (if any) is likely to be weak. With these insights, we analyze the ISCCPÐD2 cloud data sets to study if the expected different correlations between GCR fluxes and low, middle, and high cloudiness exist. ISCCP-D2 data sets are considered as the most reliable measure of global cloud cover30 and are widely used for diagnostic studies of the climate system31 as well as verification of climate model simulations32. We look into the infra-red (IR) cloud data because they provide spatially and temporal unrestricted measurements that include clouds over the entire globe during both day and night6,7. Figure 5 shows the global average monthly mean anomalies of (a) high, (b) middle, and (c) low IR cloud cover during last solar cycle. To smooth out the seasonal variations, the monthly anomalies are calculated by subtracting the climatic monthly average from each month on an equal area grid before averaging over the globe6,7. The variations of GCR fluxes as measured from CLIMAX (normalized to May, 1965) are also indicated in each panel (dot-dashed lines). It is very clear that the low cloud anomalies highly co-vary with the change of GCR fluxes as has been reported by Marsh and Svensmark6,7. During one solar cycle, the absolute amount of low cloudiness changes by ~1.5-2%. The fluctuation of middle cloud anomalies is small compared to that of low cloud, and no obvious correlation exists between middle cloudiness and GCR variations. For the high cloud anomalies, there is no obvious correlation for the whole solar cycle. There may have several explanations for this. First, it takes much longer time for new particles to grow to the size of CCN or ice nuclei (IN) in high altitude than in low altitude due to much lower precursor vapor concentrations. As a result, the initial difference in CN production rate may not lead to obvious difference in CCN/IN abundance as a result of coagulation, scavenging, and mixing. Second, the properties of high cloud are determined by ice nuclei abundance which may be insensitive to CN production rate. The processes controlling IN abundance in high altitude are currently not well known. Third, there may exit a negative correlation but it does not appear in the ISCCP-D2 data of last solar cycle because of the influence of other processes such as volcano eruptions and El Niño events. We note that there were two major volcano eruptions during the period (El Chichón in April 1982 and Mt Pinatubo in June 1991). Volcano eruptions can inject large amount of SO2 into the stratosphere which leads to the formation of sulfate aerosols. On one hand, the cooling of upper troposphere as a result of volcano eruption may enhance the high cloud formation. On the other hand, the volcano aerosols descending from the stratosphere to the upper troposphere are likely to increase the frequency and lifetime of cirrus clouds33-35 and hence high cloudiness. The timescale to disperse the volcanic stratospheric aerosols around the whole globe through meridional circulations is 1-2 years36-38. Therefore, the effect of volcanic eruptions on global high cloudiness may become obvious 1-2 years after the eruptions. This is consistent with the observed increase in high cloudiness in 1984 and 1993 (i.e., 1-2 years after the El Chichón and Mt Pinatubo eruptions). A detailed analysis of Stratospheric Aerosol and Gas Experiment (SAGE) I and II aerosol extinction data for the upper troposphere39,40 indicates that a substantial enhancement of aerosols down to 2-3 km below the tropopause persisted until 1986 for the El Chichón eruption (i.e., ~4 years after the eruptions). The high cloudiness in 1987 may have been affected by the El Niño event during that year41. El Chichón Mt Pinatubo 1.5 High cloud GCR flux variations (%) 1 0 0.5 -10 0 -0.5 -20 -1 Change in cloudiness (%) -1.5 (a) -30 1.5 Middle cloud 1 0 GCR flux variations (%) 0.5 -10 0 -0.5 -20 -1 Change in cloudiness (%) -1.5 (b) -30 1.5 GCR flux variations (%) 1 Low cloud 0 0.5 -10 0 -0.5 -20 -1 Change in cloudiness (%) -1.5 (c) -30 1982 1984 1986 1988 1990 1992 1994 1996 Year Figure 5. The global average monthly mean anomalies of (a) high, (b) middle, and (c) low IR cloud cover during last solar cycle. The variations of galactic cosmic ray (GCR) fluxes as measured from CLIMAX (normalized to May, 1965) are also indicated in each panel (dot-dashed lines). The shaded areas in Figure 2(a) corresponding to the years that global high cloudiness might have been affected by volcano eruptions and El Niño event. The shaded areas in Figure 5(a) corresponding to the years that global high cloudiness might have been affected by volcano eruptions and El Niño event. From 1988 to 1993, the impact of volcano eruptions and El Niño on global high clouds is likely negligible, and it is during this period that we find a significant anti-correlation between GCR fluxes and high cloud anomalies. The increase of high cloudiness during 1988-1989 and the decrease of high cloudiness during 1991-1992 can be readily explained by the potential role of GCR in aerosol formation and CCN abundance. Furthermore, if we look at the whole period from 1984 to 1994, we do not see obvious enhancements of high cloudiness over average values due to volcano eruptions. If we consider what the volcano eruptions might have superimposed on the natural variable high cloudiness, it is clear that volcano eruptions might have enhanced the global high cloudiness by up to ~ 1.5% (without volcano effect, the high cloudiness during the solar minimum 1986-1987 is expected to be 1.5-2% less the value during solar maximum 1990-1992). In summary, the predicted different sensitivities of the particle production to cosmic ray changes at different altitudes seem to be consistent with the observed different correlations between cosmic ray variations and low, middle and high cloud anomalies. However, due to the limit of cloud cover data available and uncertainties in the volcano and El Niño impacts, our conclusions, especially with regard to the existence of anti-correlation between high cloudiness and cosmic ray variations, are not definitive. More research is obviously needed. 5. CLIMATE IMPLICATIONS While the first key step of GCR-CN-CCN-Cloud hypothesis seems to be consistent with the observed spatially-dependent correlations of GCR variations and global cloudiness, much more work is needed to clearly establish the GCR-Cloud connection. Nevertheless, it is meaningful to discuss the climate implications associated with the possible GCR-induced cloud changes. We assume that the anomalies of high cloud cover correlate negatively while that of low cloud cover correlate positively with GCR variations, and the magnitudes of the fluctuations are similar (1.5- 2% absolute change). As a result of opposite systematic variations of low and high clouds associated with solar activity, the total global cloud cover may show no obvious correlation with GCR variations. However, the radiative effects are unlikely to cancel each other. First, the net radiative forcing of clouds depends on their altitude and optical thickness. High optically thin clouds tend to warm while optically thick high and low clouds tend to cool2. Since cloud plays an import role in the Earth radiation budget, a systematic absolute increase of high cloud amount by ~1.5-2% and a decrease of low cloud amount by ~1.5-2% from solar minimum to solar maximum, if confirmed, may represent an important mechanism to amplify the effect of solar variability on Earth’s climate. Second, the opposite change in high and low clouds may change the atmosphere heating profile and the distribution of energy between the atmosphere and the surface, and hence may have far-reaching dynamical and climatic consequences. A systematic increase in high cloud may either warm or cool the atmosphere and the earth’s surface below, depending on the types of high clouds and the underlying atmospheric properties. For example, it has been shown that the presence of a cirrostratus (with a base height of 16 km and thickness of 1.5 km) in an otherwise clear tropical atmosphere has a net cooling effect for the atmosphere below ~6 km but has a net heating effect for the atmosphere above ~ 6 km when solar zenith angles are small (< ~ 60o)42. A systematic decrease in low cloud is likely to warm the surface by allowing more sunlight to reach the earth surface, while the same decrease will cool the lower troposphere by reducing the visible absorption in the cloud layer and infra-red absorption in the cloud layer and the atmosphere below42. The long-term trend of global low and high cloud cover as a result of GCR variations may become an important external forcing of Earth climate system. Based on observations, Lockwood et al43 have shown that from 1964 to 1996 the strength of the solar magnetic flux, shielding the Earth from GCR, has increased by ~ 41% while GCR has decreased by ~3.7%43. The ion chamber measurements44 also indicate that the sea level GCR intensity has decreased by ~2% from 1979 to 1994. The GCR intensity decrease is expected to be larger at higher altitudes in the troposphere. From the data available, we estimate that the decrease in GCR fluxes during the past two decades (1979-1999) is about 1/3 to 1/2 of the maximum variations during the last solar cycle. Thus, if the connection between low and high cloudiness exists, the global mean low cloud amount might have been decreasing (0.25-0.5% per decade) and high cloud amount increasing (0.25-0.5% per decade) during the past two decades. The net impact of GCR variations during the past two decades are likely to have warmed the earth surface but cooled the lower troposphere. Note that the potential GCR-induced change in cloud albedo and absorption may enhance such an impact (a decrease in cloud droplet concentration due to fewer CCN as a result of reduced GCR fluxes may also imply a low cloud albedo and absorption). While the exact amount of net radiative forcing associated with GCR-induced low and high cloud changes remains to be investigated, it is physically plausible that the decrease in GCR fluxes during the past two decades has led to a net warming of ~0.05 oC per decade at the surface while a net cooling of ~0.05 oC per decade in the lowest 8 km of atmosphere. In other words, the GCR-induced low and high cloud changes may explain why the Earth’s surface has warmed much more rapidly (~0.1 oC per decade) than the lowest 8 km of atmosphere during the last two decades. A piece of suggestive evidence to support such a claim is that the regions of large differences in surface and low atmospheric temperature trends remarkably coincide with the regions of high correlation between cosmic ray and low cloud top temperature as shown in Figures 6 and 7. The spatial correlation map in Figure 6 shows how ISCCP-D2 low cloud top temperatures co-vary with GCR flux. The correlation coefficients, r, are calculated from the 12-month running mean at each grid point (following Marsh and Svensmark 6,7). Figure 6 reveals a band of significantly high correlation centered around the tropics where stratocumulus and marine stratus clouds are dominant6,7. The fraction of Earth with r > 0.6 is ~ 30% which is significant at 99.9%. Figure 7 shows the global distribution of temperature trends in the lowest five miles of the atmosphere derived from MSU t2lt data for the period 1979-2000. If we compare MSU t2lt trends with similar temperature trends derived from surface observations (not shown), we see warming over the northern third of the globe both at the surface and in the five-mile-deep layer of air above. The largest differences in the trends of temperature between surface and low atmosphere measurements are over the tropical regions where the surface data show a significant warming while MSU t2lt data show a slightly cooling. The tropical radiosonde temperature data show the same patterns of surface warming and tropospheric cooling since 1979 as the independent surface and MSU observations16. As we have mentioned, the regions of large differences in surface and low atmospheric temperature trends remarkably coincide with the regions of high correlation between cosmic ray and low cloud top temperature (Figure 6). Such a nice coincidence may suggest that the differences in surface and low atmospheric temperature trends over tropics are associated with the solar indirect forcing via GCR-Cloud link. It may become necessary to include the solar indirect forcing via GCR-induced cloud change in the future climate model, as current models still cannot account fully for the apparent difference between observed surface and troposphere temperature trends since 1979 13-15. Unlike homogeneous greenhouse gases which warm both the surface and low troposphere, the potential influence of GCR variations on clouds are different at different regions of the atmosphere and the associated radiative forcing are spatial and temporal inhomogeneous. The observed rapid warming during the past two decades over the northern third of the globe both at the surface and in the air above is likely due to the greenhouse effect. Figure 6. Global correlation map of GCR with anomalies of IR low cloud top temperature. White pixels indicate regions with either no data or an incomplete monthly time series. Figure 7. Low-to-middle atmosphere decadal linear temperature trend ( oC/decade) for period 1979-2000 derived from MSU t2lt data. 6. SUMMARY AND DISCUSSION The dependence of ultrafine production rate on galactic cosmic ray ionization rate at different altitudes has been investigated. Our primary studies indicate that an increase in GCR ionization rate leads to an increase in CN production in the lower troposphere (>680 mb), but a decrease in CN production in the upper troposphere (<440 mb). In the lower troposphere, the ionization rate is low and the H2SO4 concentration is relatively high, the particle formation is limited by ionization and an increase in ionization rate leads to an increase in nucleation. In the upper troposphere, the ionization rate is very high and the H2SO4 concentration is relatively low, the particle formation is limited by H2SO4 concentration and an increase in ionization rate inhibit the nucleation by reducing the lifetime of ion clusters. The average change of CN production as the ionization rate increases is small in the middle troposphere (440-680 mb). Since an increase in ultrafine production rate is likely to increase the CCN abundance and cloudiness, we can expect that the correlation between GCR changes and global cloud cover (if any) should be positive for low cloud, negative for high cloud, and weak for the middle cloud. In addition to the reported positive correlation between GCR variations and low cloudiness, our analyses of ISCCP D2 IR cloud data further reveal that high cloudiness may be anti-correlated with GCR variations if volcano and El Niño impacts are excluded. During a solar cycle, the absolute change of high and low cloud amounts is opposite in sign but similar in magnitude (~1.5-2%). The fluctuations of middle cloud anomalies are small compared to that of low clouds, and no obvious correlation exists between middle cloudiness and GCR variations. Therefore, the observed different correlations between GCR variations and low, middle and high cloud anomalies seem to be consistent with the predicted dependence of CN production on GCR variations at different altitudes. Such a consistency suggests that solar activity might affect global cloudiness by modulating GCR fluxes. However, due to the limit of cloud cover data available and uncertainties in the volcano and El Niño impacts, our conclusions, especially with regard to the existence of anti-correlation between high cloudiness and cosmic ray variations, are not definitive. The climate implications associated with the possible GCR-induced cloud changes are discussed. Since cloud is critical to Earth radiation budget, opposite systematic variations of low and high clouds associated with solar activity, if confirmed, may represent an important mechanism to amplify the effect of solar variability on Earth’s climate. The decrease in GCR intensity during the last two decades might have led to a decrease in global mean low cloud amount and an increase in high cloud amount, which might have warmed the Earth’s surface and cooled the low troposphere. The potential GCR-induced change in cloud albedo and absorption may enhance such an impact (a decrease in cloud droplet concentration due to fewer CCN as a result of reduced GCR fluxes may also imply a low cloud albedo and absorption). We suggest that, the GCR-induced natural variability of global cloudiness, together with the greenhouse gases which warm both the surface and low troposphere, may reconcile the apparent differences in global mean temperature trends at Earth’s surface (rapidly warming, as recorded by thermometers) and in the lowest 8 km of atmosphere (little if any warming, as monitored by satellites and balloons). While this study provides additional evidence for the inferred correlation between variations in global cloud properties and the solar-modulated GCR fluxes, much more work is needed to understand how and how much the GCR variations will affect the global cloud properties. The first key process (i.e, influence of GCR variations on nucleation and CN abundance) in our proposed GCR-CN-CCN-Cloud hypothesis seems to be consistent the spatially dependent influence of GCR variations on cloud properties. However, we currently do not know how much the natural GCR variations will affect the CCN abundance and cloud properties. Laboratory and field measurements, as well as theoretical studies are needed to validate the predicted dependent- behaviors of nucleation on ionization rates at different altitudes, to investigate the effect of GCR variations on CCN abundance, and to clarify the complex microphysics of aerosol/cloud interactions. The current analyses of GCR-cloud correlations are limited by the uncertainties associated with the cloud data and short periods of cloud data available. Improved cloud cover data covering longer time periods will be very useful in studying GCR-cloud connections. ACKNOWLEDGEMENTS This work was supported by NSF grant and start-up fund from State University of New York, Albany. REFERENCES 1. IPCC, Climate Change 1995: The Science of Climate Change, Intergovernmental Panel on Climate Change. Ed. by J. T. Houghton et al., Cambridge Univ. Press, New York (1996). 2. Hartmann, D.L. Radiative effects of clouds on Earth’s climate, in Aerosol-Cloud-Climate Interactions. Ed. by P.V. Hobbs, Academic Press Inc., San Diego (1993). 3. Svensmark, H., and E. Friis-Christensen, Variation of cosmic ray flux and global cloud coverageÐA missing link in solar-climate relationships. J. Atmos. Sol. Terr. Phys., 59, 1225 (1997). 4. Rossow, W.B., and R.A. Schiffer, ISCCP Cloud Data Products. Bull. Am. Met. Soc., 72, 2- 20 (1991). 5. Rossow, W.B., and R.A. Schiffer, Advances in Understanding Clouds From ISCCP. Bull. Amer. Meteor. Soc., 80, 2261-2287 (1999). 6. Marsh, N., and H. Svensmark, Cosmic rays, clouds, and climate. Space Sci. Rev., 94, 215- 230 (2000a). 7. Marsh, N., and H. Svensmark, Low Cloud Properties influenced by Cosmic Rays, Phys. Rev. Lett., 85, 5004-5007 (2000b). 8. Kernthaler, S. C., R. Toumi, and J. D. Haigh, Some doubts concerning a link between cosmic ray fluxes and global cloudiness. Geophys. Res. Lett., 26, 863-865 (1999). 9. J¿rgensen, T. S., and A. W. Hansen, Comments on “Variation of cosmic ray flux and global cloud coverage Ð a missing link in solar-climate relationships” by Henrik Svensmark and Eigil Friis-Christensen. J. Atmos. Sol. Terr. Phys., 62, 73-77 (2000). 10. Kristjánsson, J. E., and J. Kristiansen, Is there a cosmic ray signal in recent variations in global cloudiness and cloud radiative forcing? J. Geophys. Res., 105, 11,851-11863 (2000). 11. IPCC, Climate Change 2001: The Scientific Basis, Working Group I contribution to the IPCC Third Assessment Report: Summary for policymakers. Intergovernmental Panel on Climate Change, Shanghai Draft 21-01-2001 (2001). 12. NRC, Reconciling Observations of Global Temperature Change. Panel on Reconciling Temperature Observations, National Research Council, National Academy Press, Washinton D. C. (2000). 13. Hansen, J. M., M. Sato, and R. Ruedy, Radiative forcing and climate response. J. Geophys. Res., 102, 6831-6864 (1997). 14. Bengtsson, L., Roeckner, E., and Stendel, M., Why is the global warming proceeding much slower than expected? J. Geophys. Res., 104, 3865-76 (1999). 15. Santer, B. D., et. al., Interpreting differential temperature trends at the surface and in the lower troposphere. Science, 287, 1227-1232 (2000). 16. Gaffen, D. J., et al., Multidecadal changes in the vertical temperature structure of the tropical troposphere. Science, 287, 1242-1245 (2000). 17. Millikan, R. A., H. V. Neher, and W. H. Pickering, Further studies on the origin of cosmic rays. Phys. Rev., 66, 295-302 (1944). 18. Neher, H.V., Cosmic-Rays at high latitudes and altitudes covering four solar maxima. J. Geophys. Res., 76, 1637-1651 (1971). 19. Yu, F., and R. P. Turco, Ultrafine aerosol formation via ion-mediated nucleation. Geophys. Res. Lett., 27, 883-886 (2000). 20. Yu, F. and R. P. Turco, From molecular clusters to nanoparticles: The role of ambient ionization in tropospheric aerosol formation. J. Geophys. Res., 106, 4797-4814 (2001). 21. Weber, R. J., et al., Measurement of new particle formation and ultrafine particle growth rates at a clean continental site. J. Geophys. Res., 102, 4375-4385 (1997). 22. Yu, F., and R. P. Turco, The role of ions in the formation and evolution of particles in aircraft plumes. Geophys. Res. Lett., 24, 1927-1930 (1997). 23. Yu, F., R. P. Turco and B. Kärcher, The possible role of organics in the formation and evolution of ultrafine aircraft particles. J. Geophys. Res., 104, 4079-4087 (1999). 24. Kärcher, B., F. Yu, F. P. Schroeder and R. P. Turco, Ultrafine aerosol particles in aircraft plumes: Analysis of growth mechanisms. Geophys. Res. Lett., 25, 2793-2796 (1998). 25. Weber, R. J., et al., New particle formation in the remote troposphere: A comparison of observations at various sites. Geophys. Res. Lett., 26, 307-310 (1999). 26. Clarke, A. D., Kapustin, V. N., Eisele, F. L., Weber, R. J., and P. H. McMurry, Particle production near marine clouds: Sulfuric acid and predictions from classical binary nucleation. Geophys. Res. Lett., 26, 2425-2428 (1999). 27. Twomey, S., The influence of pollution on the shortwave albedo of clouds. J. Atmos. Sci., 34, 1149-1152 (1977). 28. Albrecht, B.A., Aerosols, cloud microphysics, and fractional cloudiness. Science, 245, 1227-1230 (1989). 29. Hobbs, P.V., Aerosol-cloud interactions. In: Aerosol-cloud-climate interactions. Ed. by P.V. Hobbs, Academic Press Inc., San Diego (1993). 30. Stubenrauch, C. J., W. B. Rossow, F. Chéruy, A. Chédin, and N.A. Scott, Clouds as seen by satellite sounders (3I) and imagers (ISCCP), Part I, Evaluation of cloud parameters. J. Clim., 12, 2189-2213 (1999). 31. Haeffelin, M., R. Kandel, and C. Stubenrauch, Improved diurnal interpolation of reflected broadband shortwave observations using ISCCP data. J. Atmos. Oceanic Technol., 16, 38- 54 (1999). 32. Gates, W. L., et al., An overview of the results of the Atmospheric Model Intercomparison Project (AMIP I), Bull. Am. Meteorol. Soc., 80, 29-55 (1999). 33. Jensen, E.J., and O. B. Toon, The potential effects of volcanic aerosols on cirrus cloud microphysics. Geophys. Res. Lett., 19, 1759-1762 (1992). 34. Knollenberg, R.G., Kelly, K., and J. C. Wilson, Measurements of high number densities of ice crystals in the tops of tropical cumulonimbus. J. Geophys. Res., 98, 8639-8664 (1993). 35. Wylie, D. P., W. P. Menzel, H. M. Woolf, and K. I. Strabala, Four years of global cirrus cloud statistics using HIRS. J. Clim., 7, 1972-1986 (1994). 36. Trepte, C. R., and M. H. Hitchman, Tropical stratospheric circulation deduced from satellite aerosol data. Nature, 355, 626Ð628 (1992). 37. Trepte, C. R., R. E. Veiga, and M. P. McCormick, The pole-ward dispersal of Mount Pinatubo volcanic aerosol. J. Geo-phys. Res., 98, 18,563Ð18,573 (1993). 38. Holton, J. R., P. H. Haynes, M. E. McIntyre, A. R. Douglass, R. B. Rood, and L. Pfister, Stratosphere-troposphere ex-change. Rev. Geophys., 33, 403Ð439 (1995). 39. Kent, G. S., Wang, P. H., McCormick, M.P., and K. M. Skeens, Multiyear stratospheric aerosol and gas experiment II measurements of upper tropospheric aerosol characteristics. J. Geophys. Res., 100, 13,875-13,899 (1995). 40. Kent, G. S., Trepte, C. R., and P. L. Lucker, Long-term stratospheric aerosol and gas experiment I and II measurements of upper tropospheric aerosol extinction. J. Geophys. Res., 103, 28,863-28,874 (1998). 41. Wang et al., A 6-year climatology of cloud occurrence frequency from Stratospheric Aerosol and Gas Experiment II observations (1985-1990). J. Geo-phys. Res., 101, 29407- 29 (1996). 42. Liou, K.N., Radiation and Cloud Processes in the Atmosphere: Theory, Observation, and Modeling. p305, 393-395, Oxford University Press, New York, 487 pp (1992). 43. Lockwood, M., R. Stamper, and M.N. Wild, A doubling of the Sun’s coronal magnetic field during the past 100 years. Nature, 399, 437-439 (1999). 44. Ahluwalia, H.S., Galactic cosmic ray intensity variations at a high latitude sea level site 1937-1994. J. Geophys. Res., 102, 24229-24236 (1997). CLOUD DROPLET GROWTH R. Bingham Rutherford Appleton Laboratory, Chilton, Didcot,Oxon, U.K. Abstract We present a new mechanism for rapid cloud droplet growth. The proposed mechanism relies on an isotropic pressure introduced as a result of shadowing between two droplets and can occur in both a charged and neutral atmosphere. We consider the possibility of enhancing the growth of cloud droplets in the presence of charged particles in particular ions produced from cosmic rays. 1. INTRODUCTION The subject of raindrop formation in the presence of charged particles is not new, in C.T.R. Wilson 1899 reported on experiments promoting the formation of raindrops. Wilson 1899 pointed out that a “slight rain-like condensation takes place in which a supply of ions has been produced by the action of Ro¨ntgen rays or other ionising agent”. The model we propose for the rapid growth in rain droplet size was originally used to explain the formation of dust particulates in the plasma etching process (Bingham & Tsytovich, 2001). In these plasma experiments dust agglomeration has been shown to be important in laboratory etching experiments (Garscadden et al. 1994) where the growth of dust is extremely rapid and is due to dust-dust attraction by a plasma or neutral bombardment force known as the shadow force (Bingham & Tsytovich, 2001). In the atmosphere charged drops are the norm and exist in an atmosphere of charged ions and electrons whose densities are enhanced possibly by cosmic rays. The growth of the droplet is then influenced by the presence of the plasma particles. The shadow force is caused by both ion and neutral atom droplet collisions. The mechanism of the shadow attraction even for droplets of the same charge is described by a relatively simple expression. In a charged atmo- sphere Debye screening significantly reduces the repulsive Coulomb force while the attractive force due to the bombardment are not affected by screening and dominate. The nature of the attractive force is due to bombardment by charge plasma particles and neutral atoms. For a single droplet in a plasma atmosphere with an isotropic distribution of particles, direct bombardment and deflection transfers no net momentum to an isolated droplet and therefore no net force acts on a single drop. Another drop at distance r shadows the flux to the first with a solid angle a2/r2 where a is the drop radius. the net momentum transfer is proportional to the solid angle, to the surface 2 area of the drop a and to the neutral and plasma pressure nkBT where n is the density of neutrals or plasma, kB is Boltzmann’s constant and T is the temperature of neutral/plasma. The force imported by the anisotropic pressure is given by a2 F + nk T a2 (1) r2 B a and attractive potential Ua given by a2 U = nk T a2 (2) a a r B where the coefficient a consists of three parts a = b + c + n (3) b is due to direct plasma bombardment, c is due to plasma particle screening and n is due to neutral particle bombardment (Bingham & Tsytovich, 2001). The presence of the plasma creates a flux of charged particles which results in charging as well as electron and ion recombination on the surface resulting in deposition of plasma material on the droplet. The momentum is mainly transferred from the ions and neutral atoms because of their greater mass than the electrons, and the result depends on the ion and neutral attachment coefficient (Bingham & Tsytovich, 2001). Charged particles of similar sign also produces a repulsive force which in the absence of screening 2 2 has a potential Uc = Zd c /r where Zd is the charge on the particles. It is perfectly possible for both signs to exist on the particles this would of course lead to attraction of opposites. The force given by Eq.(1) operates in the presence or absence of plasmas. In the presence of plasma the force due to the ion and neutral flux should be compared to the Coulomb force Fc between two particles given by 2 2 2 2 2 2 2 Zd e Zde nikBTe a 2 a 2 Fc = 2 = 2 2 = z 2 nikBTe4 de (4) r aTe ! e nir r 2 Zde Te where z = is the dimensionless charge of order 1-2 and d = 2 is the electron Debye radius. aTe e 4 nee 2 2 q 4dez Comparison of Eqs.(1)-(2) demonstrates that the bombardment force due to ions is a2 less than the Coulomb repulsive force for r d . For r d the Coulomb force is screened while the attractive e e shadow force is not, this can lead to a contraction of the cloud until the inter-dust distances are compa- rable to 10 times the Debye radius. For distances larger than the Debye radius the ion accretion force can dominate Coulomb repulsion and should be added to the attraction due to gas particle bombardment but this contribution is usually small. Using r 4d where the repulsive force can be overcome by the e attractive force forming an attractive potential well, with the potential given by U F r z2n d a2T (5) ' c i e e We find that a “droplet” is formed if 2 2 Td < Tenidea z (6) where Td is the droplet kinetic temperature not the surface temperature, and de is the electron Debye radius. The phenomenon of droplet formation due to ion or neutral bombardment has previously not been considered. It can also operate at a very low level of ionization. In some case it will serve as the main mechanism responsible for growth of the droplets. An estimate of the growth time is found by taking into account the relative number of particles in phase space with energies less than the attractive potential well dm 1 U k T d = 2m v n 3 a ; v = B d (7) dt d d d k T d m B d s d Since the force acting is in the direction separating the two particle we can use for the time scale the time for the particles to travel the interparticle distance which according to Eq.(4) is given by 3/2 1 2 2 1 kBTd 3 kBTenia deZ nd (8) ' s md Td ! 2 where Z = Zde /akBTe. This model of droplet formation i.e. the coalescence of successive scale sizes of droplets produces large drops at a much faster rate than other processes, such as the continuous drop model. The model proposed in this paper is inherently stochastic in nature. The shadow force introduced in this paper to enhance the growth of droplets by coalescence or agglomeration has as far as we know not been used to calculate the growth rate. It is obvious from the model that the presence of ions enhances the process. The link with cosmic rays through ion production is obvious and therefore an enhance cosmic ray flux would lead to enhanced droplet formation. Consequences of this work to the envisaged CLOUD experiment. The next stage is to develop a numerical modelling procedure which can handle a large number of coalescing droplets. References [1] C T R Wilson, Proc. Camb. Phil. Soc.11, 52, 1899 [2] R Bingham & V N Tsytovich, IEEE T Plasma Sci., 29, (2), 158-163, 2001, [3] Garscadden, A., Ganguly, B.N., Healand, P.D., & Williams, J., 1994, Plasma Sources Sci. Technol- ogy, 3, 239 ATHERMODYNAMIC-KINETIC MODEL FOR IONIC CLUSTER FORMATION, GROWTH AND NUCLEATION Raffaella D’Auria and Richard P. Turco Department of Atmospheric Sciences, University of California, Los Angeles Abstract Anew model is presented for ionic cluster formation and evolution based on the chemical kinetics of charged molecular aggregates, under thermodynamic constraints. Basic laboratory measurements of individual cluster enthalpies and entropies of formation are combined with detailed quantum mechanical calculations of ion-aggregate structures and energetics to extend the database for the smaller clusters, while the classical Thomson charged-droplet model is used to define the properties of the larger species. The result is a novel hybrid model that spans the entire range of cluster sizes. The corresponding kinetic growth equations predict cluster populations for a broad range of environmen- tal conditions. The present hybrid thermodynamic/kinetic model is applied to study large ion species and their effects in the winter polar stratosphere. 1. INTRODUCTION The formation of new particles in the atmosphere, as well as phase changes in pre-existing aerosols, has been a central topic of inquiry in a number of fields for more than a century. The problem is elucidated in the original work of J. J. Thomson, who was concerned with the nucleation of new particles from a supersaturated vapor [1]. Thomson proposed a useful representation for nucleation embryos in the form of a microscopic droplet in quasi-equilibrium with its surroundings. If such an embryo reached a “critical” size–defined by environmental conditions (such as air temperature, and the partial pressure of the condensing vapor)–it would begin to grow freely. The size of the critical embryo for nucleation, according to Thomson’s treatment, is a relatively simple function of the bulk properties of the condensed material, particularly the surface tension and vapor pressure. In Thomson’s model, and in nature, subcritical molecular clusters formed via random collisions tend to evaporate quickly even if the ambient vapor is saturated with respect to the bulk condensate. This barrier to nucleation is a manifestation of the excess energy required to build an interface isolating the embryo from its surrounding vapor. In absence of preferential sites for condensation (for example, exter- nal surfaces, or ions), nucleation is said to be “homogeneous”, and proceeds only when the environment is highly supersaturated. In the presence of energetically favorable condensation sites, however, the nu- cleation process is called “heterogeneous”, and usually proceeds under less stringent conditions (i.e.,at alower supersaturation). Ions are known to act as condensation centers. The Wilson cloud chamber is the most notable example of ion-induced condensation. The electric field associated with an ion polarizes molecules in its vicinity, creating a significant charge-dipole attractive force. Ion-molecule reactions, for example, are greatly accelerated by this attraction (typically, by more than an order of magnitude compared to normal molecular reactions). Furthermore, within an ion-molecule aggregate, the central field stabilizes the cluster, thereby reducing its tendency to evaporate. In the lower stratosphere and upper troposphere, ions are continuously formed by the deposition of galactic cosmic radiation. The cosmic ray flux decreases with increasing solar activity, and varies with geomagnetic latitude, being more intense toward the poles. The resulting ionization rate exhibits similar temporal and spatial modulation. However, for the present study, which focuses on altitudes from about 10 to 25 km, a constant mean ion-pair production-rate of 10 ion-pairs/cm−3 · s−1 can be assumed. According to the ion reaction scheme proposed by Ferguson and coworkers [2], the initial ion- ization products in air are free electrons and simple positive ions, such as N+ and O+.Below80 km or so, the electrons rapidly attach to oxygen molecules, forming negative ions. The resulting plasma of positive and negative ions, in equal concentrations, subsequently evolves through a complex series of fast ion-molecule reactions. The ions become progressively more stable, forming the commonly observed + + − − core species such as H3O , NH4 , NO3 , and HSO4 . The ultimate sink for ionization is charge recom- bination, either via ion-ion recombination, or uptake on aerosols. Under typical atmospheric conditions, ion lifetimes are long enough to allow extensive association and switching of neutral molecules with the stable core species, leading to massive charged clusters. + · Positive ions are quickly converted to proton hydrates (PHs) in the series, H (H2O)n [3], owing to high concentrations of water vapor in air. The PHs can react further with trace substances having larger proton affinities than water (indicated here as X), being transformed into the so-called non-proton + · · hydrates (NPHs). These ions are characterized by the general structure H Xn (H2O)m.For example, in the lower stratosphere at mid-latitudes, high-resolution mass spectrometry has revealed ions contain- ing acetonitrile, CH3CN (e.g., [4]). In the upper troposphere, other trace species can play a major role. Among the positive ions identified through mass spectrometric measurements are ammonium/water clus- + · ters, NH4 (H2O)n,and pyridine and acetone-containing clusters. It is expected that many other trace species will ultimately be found on large charged aggregates [5], [6]. − Among the most common negative ions are clusters of NO3 with water and/or nitric acid, and − of HSO4 with water and sulfuric acid [7], [8]. The nitrate anion is the first highly stable negative ion generated by cosmic ray ionization, because of the abundance of HNO3,N2O5 and other nitrogen − · oxides in the atmosphere. The nitrate core ion is rapidly hydrated, leading to NO3 (H2O)n clusters. The water can be displaced by other molecules, such as SO2 [9], and HNO3 [10], [11], creating mixed − · · · aggregates such as NO3 (H2O)l (SO2)m (HNO3)n [2]. The even greater stability of bisulfate and sulfate anions causes the nitrate core ion to be displaced through reactions with sulfuric acid [8]. The ambient concentrations of sulfuric acid vapor are so low, however, that the complete conversion of nitrate to sulfate core ions is relatively slow. It is noteworthy that most of the existing in situ data concerning ion size and composition, for both positive and negative species, refers to midlatitude conditions at high altitudes. Hence, there is a need for a more detailed understanding of ion composition under a much wider range of atmospheric conditions, as discussed below. Modeling of the ion composition of the lower stratosphere and upper troposphere has been carried out in the past ([12], [13], [14], [15], [16]). Predictions are roughly consistent with observations. Such modeling is feasible because the ionization sources and chemical transformations of primary ions into terminal ions, especially those identified above, are relatively well characterized through laboratory in- vestigations (e.g., [2], [17]). On the other hand, the numbers and types of ligands attached to specific core ions have not been fully resolved for ambient atmospheric conditions. The range of altitudes at which different families of positive and negative ions are dominant (in terms of their relative abundances) is ex- pected to vary with season and latitude as a consequence of systematic variations in air temperature and composition, and possibly in ionization rate (e.g.,assimulated by Beig et al. [15], using a 2-D model). At the extremely low temperatures found at high latitudes in winter, for example, significant mod- ifications of the cluster ion population are expected, which have never been sampled or simulated. The differences may have an impact on the formation of polar stratospheric clouds [18]. With temperatures in the range of 180-220 K, and long periods of darkness, the primary ionic clusters are likely to consist of + · − · − · H3O (H2O)n, NO3 (H2O)n, and NO3 (HNO3)n. D’Auria and Turco [19] argue that sulfuric acid core ions and ligands are unlikely to be important under polar winter conditions; H2SO4 concentrations are so low (e.g., [20]) in this situation that the time required for a single sulfuric acid molecule to react with a nitrate core ion greatly exceeds the recombination lifetime of the ions. 2. KINETICS AND THERMODYNAMICS OF ION CLUSTERING The formation of a typical cluster is described by the chemical reaction, k − ∗ · f,n 1 ∗ · c0 (l)n−1 + l c0 (l)n, (1) kr,n where c0 represents the core ion of sign ∗, l identifies the vapor species that is condensing on the cluster as ligands, n an integer that indicates the number of ligands on a cluster, and kf,n−1 and kr,n represent the forward and reverse rate coefficients for reaction (1), respectively1. Considering a sequence of species with 0 In Eq. (2), Q0 is the ion production rate, and rec is the ion-ion recombination rate coefficient, which accounts for cluster depletion due to charge neutralization. While the recombination loss term actually involves the summation over all ion species of opposite charge, an average recombination coefficient is used in the present analysis, where c¯ represents the total concentration of ions (of each sign). To a good approximation Q0, rec,and hence the total ion abundance, c¯, can be taken as constants in this analysis (in reality these parameters are quasi-steady, varying on time scales that are long relative to the charge lifetime). The ligand concentration, l,isusually fixed by environmental conditions. Hence, the forward and reverse rate coefficients (at the ambient temperature and pressure of interest) are the key parameters defining the kinetics of the clustering process. Indeed, once these coefficients are known, Eqs. (2) reduce to a set of linear first-order differential equations that can be integrated numerically to determine the cluster concentrations corresponding to any stable state of the atmosphere. −6 3 −1 The parameter values assumed in the present study are: rec =1× 10 cm · s [21]; Q0 = 10 ion − pairs/cm3 · s.With these ionization and recombination parameters, the steady-state ion con- centration (of each sign), c¯,isabout 3 × 103 cm−3,avalue in agreement with observations [22], [23], [24]. Moreover, the mean cluster lifetime is roughly ∼ 300 sinthis case. Note that in this derivation, the ligand vapor concentration is assumed to be insensitive to ion reactions, which is reasonable for typ- ical atmospheric ion abundances, time scales of interest, and ligand (e.g.,water and nitric acid) partial pressures. The forward and reverse rate coefficients for a clustering reaction are intimately connected to the equilibrium constant for the process. An example is given by Eq. (1), kf,n−1 [cn] Kn−1,n = = , (3) kr,n [cn−1][l] where Kn−1,n, the equilibrium constant for process (1), is also related to the cluster and ligand concen- trations through the last term in Eq. (3). The equilibrium constant is defined in terms of the Gibbs free 0 energy change, ∆Gn−1,n,associated with the clustering reaction (1), 0 ∆Gn−1,n K − = exp − , (4) n 1,n RT 1 ∗ For ease of notation, the cluster c0 · (l)n will be simply designated as cn. where the energy change corresponds to standard conditions (i.e.,ataligand partial pressure of 1 atm). Here, R is the gas constant, and T is the ambient temperature. Likewise, the standard Gibbs free energy can be written as, 0 0 − 0 ∆Gn−1,n =∆Hn−1,n T ∆Sn−1,n , (5) 0 0 where ∆Hn−1,n and ∆Sn−1,n are, respectively, the standard enthalpy and entropy for reaction (1). Once 0 these latter values are determined–experimentally or theoretically– ∆Gn−1,n can be calculated at any 0 0 temperature using Eq. (5). The values of ∆Hn−1,n and ∆Sn−1,n are fundamental properties of the clusters. Hence, so is the equilibrium constant (through Eqs. (4) and (5)). Accordingly, if kf,n−1 were measured or estimated independently, kr,n would follow immediately from Eq. (3), and the system (2) could be integrated. In the present study the forward rate coefficients, kf , are based on laboratory measurements. In this case, we assume a constant value of 1 × 10−9 cm3 · s−1 for all ligand addition reactions [25]. The forward coefficient is determined by collisional parameters (molecular speed, impact cross sec- tion) as well as steric and surface factors (which may be summarized in the form of an accommodation coefficient). Typically, forward rate coefficients for energetically favorable reactions are greater than 1 × 10−10 cm3 · s−1. The actual value does not affect the predicted equilibrium state, however, since the kinetics are constrained by Eq. (3). On the other hand, the approach of the cluster size distribution to equilibrium (and steady-state) during the course of an ion lifetime could be influenced by forward rate coefficients smaller than ∼ 10−10 cm3 · s−1. 3. SOURCES OF THERMODYNAMIC DATA TO BUILD A HYBRID MODEL In the present work, a “hybrid” approach has been taken in determining the thermodynamic parameters needed to solve the ion cluster kinetic Eqs. (2). The model incorporates enthalpy and entropy measure- ments for the clustering of ligands about core ions. Such data are typically available for clusters ranging from several to 10 or so ligands. In the case of clusters for which measurements have not been made, two approaches are taken. First, we follow the procedure described by Castleman and coworkers ([26], [27], [28]), in which the standard Gibbs free energy is estimated based on Thomson’s treatment of the ener- getics of a charged liquid sphere [1]. Second, as means of confirming the convergence of the Thomson model to laboratory measurements on the one hand, and as an independent source of new thermodynamic data on the other, we carry out quantum mechanical simulations of cluster structures and energies, from which thermodynamic parameters can be derived. This hybrid approach for extending the data is applied to several key ionic cluster sequences (i.e.,hydronium/water, nitrate/water, and nitrate/nitric acid) in the following sections. 3.1 Laboratory measurements 3.1.1 Proton hydrates There are a number of laboratory measurements of the hydronium/water (or proton hydrate, PH) cluster series. We focus here on those data sets spanning a significant range of cluster sizes (number of ligands per cluster), for which both enthalpy and entropy data have been recorded. The available measurements are summarized in Figure 1. A major source of data used in building our hybrid PH model derives from the work of Kebarle and coworkers ([29], [30], [31]). Data from Meot-Ner and Speller [32] are in good agreement with the results of Kebarle and coworkers. Because of the greater cluster size range, and the self consistency among several studies carried out by Kebarle’s group, their data are adopted as our baseline PH model (see below). The se- lected enthalpies and entropies (as well as those in [32]) are derived from van t’Hoff plots of equilibrium constants measured at different temperatures. 40 90 Data from Ref. [29] Data from Ref. [29] Data from Ref. [30] Data from Ref. [30] 35 Data from Ref. [31] 80 Data from Ref. [31] Data from Ref. [32] Data from Ref. [32] Data from Ref. [33] 70 Data from Ref. [33] 30 Data from Ref. [36] K)] 60 25 [kcal/mol] 50 [cal/(mol 20 0 n−1,n 0 n−1,n H 40 S D D − 15 − 30 10 20 5 10 0 5 10 15 20 25 30 0 5 10 15 20 25 30 n, # of H O ligands per cluster n, # of H O ligands per cluster 2 2 Fig. 1: Enthalpy change (left panel) and entropy change (right panel) associated with the clustering of water about the hydronium ion. Data are taken from the sources indicated in the figure legend. By contrast, the thermodynamic data presented by Shi et al. [33] were obtained by recording the decay fractions of metastable species at one temperature, and employing Klots’ model [34] of evaporative dissociation to assess binding energies. The enthalpy was derived from the binding energy, while the entropy was estimated using earlier data from the same group [35]. In comparing the various data sets in Figure 1, the Shi et al. results disagree significantly with the other (direct) measurements for cluster sizes from n =4−7.This disparity, together with the indirect nature of Shi et al.’s enthalpy and entropy measurements, has lead us to discount these data in building the hybrid model. Enthalpy data from Magnera et al. [36] are shown in the left-hand panel of Figure 1. These values were also derived from estimated binding energies for the PH clusters based on measurements of collision-induced dissociation in a triple-quadrupole mass spectrometer. The accuracy of the results is stated to be ± 10% for the smaller clusters, and about ± 20% for the larger ones. In the region where n>5, the enthalpy values fall into the lower end of the range established by other studies. Since these measurements were carried out at a single temperature, entropies were not reported, rendering the data of less use for our purposes. 3.1.2 Nitrate ions For nitrate/water clusters, measurements are not as extensive as for the hydronium/water system. The most complete set of enthalpy and entropy data are available from the measurements of Castleman and coworkers [37]. These data are reported in the first section of Table 1. Note that the results extend only to n =3. For the nitrate/nitric acid cluster series, the most comprehensive measurements are those of David- son et al. [38] and Wlodek et al. [39]. Davidson and coworkers reported enthalpy and entropy values for − the first three HNO3 ligands on a NO3 core ion (Table 1). Note that for n =1, the enthalpy and entropy changes were derived using the lower limit for K0,1, since the corresponding clustering reaction was not actually observed at equilibrium. The measurements of Wlodek and coworkers are also summarized in Table 1. Because the entropy change for n =6is extremely low, we do not use this value to calculate cluster distributions; the entropy is instead estimated from the Thomson model (see below). A few other limited data sets exist (e.g., Arnold and coworkers, [8], [40]), but these lack simultaneous enthalpy and entropy measurements. Table 1: Thermodynamic data for nitrate/water from [37], and for nitrate/nitric acid from [38] (center column) and [39] (right- hand column). − · − · NO3 (H2O)n NO3 (HNO3)n − − 0 − 0 − 0 − 0 − 0 − 0 n 1,n ∆Hn−1,n ∆Sn−1,n ∆Hn−1,n ∆Sn−1,n ∆Hn−1,n ∆Sn−1,n [kcal/mol] [cal/mol·K] [kcal/mol] [cal/mol·K] [kcal/mol] [cal/mol·K] (0,1) 14.6 ± 0.225.0 ± 0.4 26 ≥ 20 -- (1,2) 14.3 ± 0.230.3 ± 0.5 18.3 ± 1.022.1 ± 2 16.0 ± 0.823.1 ± 2.4 (2,3) 13.8 ± 0.433.2 ± 1.3 16.1 ± 1.028.9 ± 2 13.9 ± 1.426.7 ± 4.4 (3,4) -- --9.3 ± 1.319.9 ± 3.4 (4,5) -- --7.4 ± 1.218.6 ± 5.0 (5,6) -- --4.6 ± 0.97.3 ± 5.0 3.2 The Thomson model The Thomson model [1] provides a convenient description for ionic clusters, attributing to them the prop- erties of small charged droplets. The effects of surface tension and curvature, and charge polarization, are explicitly included in the Gibbs free energy of the droplet. Moreover, this energy is conveniently characterized by the macroscopic properties of the condensed phase, such as surface tension, σ,bulk density, ρ, dielectric constant, r, and vapor saturation ratio, S. When the droplet is taken to be spherical with radius, r, the Gibbs free energy change associated with its formation is given by, ∆G = −nRT ln S +4πr2σN + (6) 0,n A N q2 1 1 1 + A 1 − − , 8π 0 r r r0 The saturation ratio, S,isdefined as the quotient of the ligand partial pressure and the saturation vapor pressure above a flat surface of the condensed ligand material. Likewise, r is the relative dielectric constant of the condensed ligand phase. The radius of the core ion is taken to be r0, the charge of the ion is q, 0 is the dielectric constant in vacuum and NA is Avogadro’s number. The first term in Eq. (6) expresses the energy associated with the condensation of n molecules from the vapor phase into the liquid phase. The second term accounts for the work needed to construct the spherical interface for a droplet of radius r. The last term represents the Born energy corresponding to the solvation of an ion in the liquid condensate. Note that r and n are equivalent variables in this representation, and are connected through the ligand density and molecular weight. From Eq. (6), with the ligand partial pressure fixed at 1 atm (for “standard conditions”), the free energy change associated with clustering step n is computed as, 0 0 − 0 ∆Gn−1,n =∆G0,n ∆G0,n−1 . (7) The entropy change then follows from Eq. (5), by differentiating Eq. (7) with respect to temperature, 0 0 ∂∆Gn−1,n ∆S − = − . (8) n 1,n ∂T p=1 atm In order to calculate the entropy change in this way, the temperature dependences of the parameters, ρ, σ, and r, must be available over the temperature range of interest (and for the relevant compositions when the system is not homogeneous). Hence, detailed laboratory measurements of these macroscopic properties at different temperatures can lead to a satisfactory estimate of the thermodynamic parameters for the larger clusters, although such data are not always available, or easy to obtain. Once the entropy change has been determined, the corresponding enthalpy change can be calcu- lated directly from Eq. (5). Despite the fact that the Thomson model is based on a number of significant approximations– for example, surface capillarity, and insensitivity to the sign of the central charge–this simple treatment nevertheless appears to converge to the more precise laboratory-based thermodynamic measurements at a relatively small number of ligands per cluster, on the order of ten or less (as is shown later; also see, for example, [28]). The largest deviations of the Thomson model from measurements occur for the entropies of small clusters, although the enthalpies in this case also show significant differences. Entropies are closely related to cluster structure, which is not accounted for by the Thomson model; i.e., Thomson clusters are treated as homogeneous liquid droplets. It is logical, therefore, to employ quantum mechanical molecular simulations to extend the laboratory data base into the intermediate cluster size range where the Thomson representation is less viable. 3.3 Quantum mechanical simulations As discussed in the previous section (Sec. 3.2), the Thomson model is expected to predict the thermo- dynamic properties of larger clusters, in which the condensed materials resemble the bulk phase. On the other hand, the Thomson model cannot be used directly to represent ionic clusters with a relatively small number of ligands. Laboratory data provide the best description when such information is available. However, in the intermediate cluster size range, with perhaps 5-15 ligands, laboratory data, even when published, tend to be more uncertain. There is also a dearth of measurements for many atmospheric ion/ligand systems. Hence, a limited approach restricted to laboratory measurements and the Thomson model would be feasible only in a very limited number of cases. We extend the hybrid thermodynamic model by incorporating quantum mechanical (QM) simu- lations for small and intermediate size ionic clusters. The QM predictions for the smallest clusters are initially utilized to validate computational techniques against measurements. The predictions for larger clusters are then employed to expand the thermodynamic data base into the intermediate cluster size range. For the present analysis, density functional theory (DFT) [41] is applied to calculate cluster struc- tures and energetics using a specific hybrid functional B3LYP [42] with a 6-311++G(d,p) basis set. Computationally, the simulations are carried out using the Gaussian code [43]. 3.3.1 Results for the proton hydrates A comparison between laboratory data for the hydronium/water cluster series and DFT simulations is given in Table 2. For the smallest clusters (n<4 − 5), both measurements and DFT calculations are sig- Table 2: Comparison of thermodynamic parameters for hydronium/water clusters based on laboratory data and DFT simula- tions. − − 0 − 0 n 1,n ∆Hn,n−1 [kcal/mol] ∆Sn,n−1 [kcal/mol] (0,1) 36a 31.6b - 36.99d 23.91e 33.3a 24.3b - 29.14d 25.14e (1,2) 22.3 19.5 - 22.28 15.53 29 21.7 - 25.25 24.85 (2,3) 17 17.5 17.9c 18.26 12.72 28.3 27.3 28.4c 24.92 24.94 (3,4) 15.3 - 12.7 13.06 11.4 32.6 - 23.4 25.49 25.08 (4,5) 13 - 11.6 11.75 10.67 30.3 - 25.0 26.85 25.21 (5,6) 11.7 - 10.7 10.54 10.22 29.6 - 26.2 24.28 25.33 (6,7) 10.3 - - 9.40 9.93 27 - - 25.03 25.43 a: Data in this column are from [29]. b: Data from [30]. c: Data from [31]. d:Values in this column are derived from DFT calculations (B3LYP/6-311++G**). e:Thomson model predictions. nificantly larger than the Thomson model results. Because the Thomson model incorporates macroscopic quantities, such as density and surface tension, the resolution of specific effects related to molecular con- figurations within clusters is not possible. The corresponding Gibbs free energies are illustrated in Figure 2. The Thomson model produces a rather smooth trend, as expected. It is noteworthy, however, that all of the free energy curves appear to converge as n reaches intermediate sizes (n>10)(the QM calculation for n =11in Fig. 2 is a preliminary result corresponding to only one initial molecular configuration). The free energies also approach the theoretical limit, RTlnp0,asn →∞. 30 Data from Ref. [29] Data from Ref. [30] 25 Data from Ref. [31] Data from Ref. [32] Data from Ref. [33] 20 Thomson model QM simulations 15 [kcal/mol] 0 n−1,n 10 G D − 5 0 0 1 2 3 4 5 6 7 8 9 10 n, # of H O ligands per cluster 2 Fig. 2: Gibbs free energy change associated with hydronium/water clustering as obtained from the Thomson model, DFT simulations (noted as QM simulations in the legend), and laboratory data. By n =5,the DFT result has almost converged with the Thomson model prediction. Beyond that point, the DFT energies might be expected to fall close to the Thomson curve. From the point of view of the Gibbs free energy, this agreement implies that the macroscopic Thomson approach offers a credible description for this system. The lack of convergence evident for n =11is probably associated with the need for more simulations of isomeric configurations that might contribute to the thermochemical properties of these larger, complex clusters. We have not yet carried out a sufficient number of DFT simulations to resolve this discrepancy. It has been observed by Lau et al. [31] that the decrease in cluster stability measured after + · − 0 − 0 H3O (H2O)3, namely ∆H3,4,issignificantly lower than ∆H2,3. This feature was not appar- ent in the first measurements from Kebarle’s group [29]. However, the changes in the cluster stability seen in the data of Lau et al. (together with that of Cunningham et al. [30]) is also evident in the mea- surements of Meot-Ner and Speller [32]. Lau and coworkers compared the trend in their data with the stabilization energies, ∆En−1,n, derived by Newton [44] using ab initio molecular orbital calculations (with a 4-31G basis set). The measured and calculated energy changes exhibit similar trends beyond the + cluster, H3O · (H2O)3. Our computed stabilization energies show the same behavior as the calculations of Newton, as illustrated in the left-hand panel of Figure 3. For comparison, the cluster enthalpy changes from [31], [30] and [32] are shown in the right-hand panel. In these plots, the stepwise differences (from cluster to cluster) in the configurational energy (left panel) and enthalpy (right panel) are quite comparable. Indeed, the remarkable agreement between measurements and QM calculations provides a measure of validation of the DFT approach. At n =4,for example, the monotonically decreasing trend in −∆(∆En−1,n) and −∆(∆Hn−1,n) reverses abruptly in both sets of data. This reversal in the bonding energy differences appears to correspond to the completion of the first solvation shell. However, Meot-Ner and Speller point out that, while this “shell effect” may ± 0 be real, it lies within experimental error (which they estimate as 1 [kcal/mol] for the ∆Hn−1,n values in Figure 3). Similarly, a noticeable decrease in the absolute value of entropy change occurs beyond n =3in both the Meot-Ner and Speller and combined Cunningham-Lau measurements. Meot-Ner and Speller, 40 40 From Ref. [44] Data from Ref. [30] 36 QM simulations Data from Ref. [31] 35 Data from Ref. [32] 32 30 28 25 24 [kcal/mol] 20 [kcal/mol] 20 n −1, 0 n−1,n n 16 H E 15 D D − − 12 10 8 5 4 0 0 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 n, # of H O ligands per cluster n, # of H O ligands per cluster 2 2 Fig. 3: Left panel: Configurational stabilization energies, and differences in stabilization energies, as computed by Newton [44] (with a 4-31G basis set), and in the present work (with a B3LYP/6-311++G basis set). Right panel: Enthalpies and enthalpy differences from laboratory data (refer to the legend for sources). + · Fig. 4: Optimized structures for the H3O (H2O)3 cluster. In the left panel, the structure shown is believed to + represent the first completed solvation shell; the structure in the right panel, which is H5O2 -centered, is about 3.3 [kcal/mol] less stable. Oxygen atoms are represented by red spheres, and hydrogen atoms by white spheres. by assuming a typical error in the entropy measurements of ± 2 [cal/(mol·K)], conclude that an entropy “shell effect” cannot be confirmed. Neither is an obvious effect seen in the present entropy calculations. Results from our study, and similar work by Wei and Salahub [45] show that for n =3, there are two dominant structures for the hydronium/water clusters. These are illustrated in Figure 4. One is + · ahydronium-centered H3O (H2O)3 structure, shown in the left-hand panel of Figure 4. The second + consists of a stable H5O2 core ion with two attached water molecules (right-hand panel). The structure on the left corresponds to the first complete solvation shell in this cluster series. The two structures differ + in energy by ∼ 3.3 [kcal/mol], with the H3O -centered cluster being more stable. A similar result was obtained by Wei and Salahub [45]. A straightforward Boltzmann weighting yields the relative equilibrium concentrations of these two isomers corresponding to n =3(considering other isomers to be unfavorable). Owing to the energy difference between the two stable forms, the hydronium-centered structure accounts for ≈ 95% of the clusters of this size (mass) for standard conditions. As the number of water ligands per cluster increases, the problem of identifying the minimum energy configuration over the entire ensemble of structures becomes more difficult. First, the DFT calcu- lations require substantially more computer time for each ligand added (making it impractical to sample a wide range of initial configurations). Second, the inherent “floppiness” of the weakly hydrogen-bonded hydronium/water clusters allows many potentially stable configurations to be selected. Accordingly, as noted in discussing Figure 2, we consider the present simulations at the largest cluster sizes to be prelim- inary. We are expanding the QM treatment to include molecular dynamics and Monte Carlo approaches to handle the large ensemble of structures needed to study the most massive clusters. 3.2.2 Results for the nitrate ions Acomparison of laboratory data, Thomson model results, and DFT calculations for the enthalpies and entropies of nitrate/water clusters is given in Figure 5. The present results for these clusters is tentative, however, and additional laboratory and structural information is needed. From the optimized configu- rations derived to date, the formation of a solvation shell for n>4 is predicted. This solvation effect is indicated in Figure 5. However, beyond n =3, further structural optimizations are needed to assess the thermodynamic properties of the clusters, and to assure convergence to the Thomson model. It is apparent, for example, that neither the Thomson predictions or QM calculations agree with the limited (n<4) laboratory measurements beyond the first cluster. This discrepancy may be a consequence of the difficulty in isolating these complex species under laboratory conditions, or a result of faulty theoretical assumptions. The matter remains to be resolved. 15 34 Data from Ref. [37] Data from Ref. [37] Thomson model Thomson model 14 QM simulations 32 QM simulations 13 K)] 30 12 [kcal/mol] 11 28 [cal/(mol 0 n−1,n 0 n−1,n H 10 S D D − 26 − 9 24 8 7 22 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 n, # of H O ligands per cluster n, # of H O ligands per cluster 2 2 − Fig. 5: Enthalpies and entropies of hydration of the NO3 ion as derived from measurements, the Thomson model, and DFT simulations. Fornitrate/nitric acid clusters, a similar comparison between thermodynamic measurements, Thom- son model predictions, and DFT simulations is given in Figure 6. For this cluster series, both measure- ments and DFT calculations appear to converge toward the Thomson model, at least with respect to the Gibbs free energy and enthalpy. Larger discrepancies occur with respect to the entropy. Typically, entropy values are obtained by differencing two large quantities–the Gibbs free energy, and the en- 0 thalpy. The resulting uncertainty in the entropy is therefore inherently greater than those in ∆Gn−1,n 0 and ∆Hn−1,n.Inthe case of the DFT simulations, the number of configurations that we have optimized at this time is probably too small to be definitive, especially for n =6.Even for the smaller clusters, additional simulations may be needed to reveal other potential minimum energy configurations. The Thomson model is not expected to provide an accurate picture of the clusters at low values of n. Indeed, in applying that model, we assumed the density, surface tension, and dielectric constant for pure nitric acid, and we estimated the derivatives (with respect to temperature) of those properties. Both of these approximations introduce additional uncertainty into the derived thermodynamic parameters. 24 33 Data from Ref. [38] Data from Ref. [38] 21 Data from Ref. [39] 30 Data from Ref. [39] Thosmson model Thosmson model QM simulations QM simulations 18 27 24 15 21 12 [kcal/mol] [kcal/mol] 18 9 0 n−1,n 0 n−1,n 15 H G D D − − 6 12 3 9 0 6 −3 3 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 n, # of HNO ligands per cluster n, # of HNO ligands per cluster 3 3 35 Data from Ref. [38] Data from Ref. [39] 30 Thosmson model QM simulations 25 K)] 20 [cal/(mol 0 n−1,n S D 15 − 10 5 0 1 2 3 4 5 6 7 8 9 10 n, # of HNO ligands per cluster 3 Fig. 6: Gibbs free energy, enthalpy, and entropy changes for the clustering of nitric acid vapor on nitrate core ions. Sources of data are identified in the legend. Despite the limitations of the present hybrid thermodynamic analysis, it is apparent that the be- havior of the clusters under investigation is bounded, and predictable, to a significant extent. Inasmuch as the enthalpy and entropy are fundamental properties of each cluster, the continuing refinement of the data base suggested above will eventually lead to an accurate, comprehensive model describing the entire spectrum of cluster sizes and compositions. 4. ION CLUSTER SIZE DISTRIBUTIONS Following the procedure described in Section 3., a set of baseline enthalpies and entropies has been generated for the ion cluster series of interest here. These data are summarized in Table 3, and are used to calculate cluster mass (size) distributions for a range of atmospheric conditions. In Section 4.3 (also, D’Auria and Turco [19]), we explore the potential effects of ion clusters on two proposed chemical and microphysical processes relevant to the stratosphere. 4.1 Evolution and relative distributions of charged clusters The system of Eqs. (2) was solved for typical upper atmospheric conditions, and the resulting evolution of the ion cluster size distributions was investigated. The characteristic time required to establish the relative abundances of the clusters varies with the ion series. For hydronium clusters, the characteristic timescale is of the order of milliseconds. The same is true of the nitrate/water cluster family (in fact, water clusters will always tend to equilibrate within milliseconds because atmospheric water vapor abundances are high). By contrast, nitrate/nitric acid ions require up to several seconds to achieve a quasi-equilibrium Table 3: Standard enthalpies and entropies for the addition of a ligand in the cluster sequences indicated, up to the first 10 ligands in each sequence. + · a − · b − · c H3O (H2O)n NO3 (H2O)n NO3 (HNO3)n − − 0 − 0 − 0 − 0 − 0 − 0 n 1,n ∆Hn−1,n ∆Sn−1,n ∆Hn−1,n ∆Sn−1,n ∆Hn−1,n ∆Sn−1,n [kcal/mol] [cal/mol·K] [kcal/mol] [cal/mol·K] [kcal/mol] [cal/mol·K] 0,1 36 33.3 14.6 25.0 26.0 ≥20. 1,2 22.3 29 14.3 30.3 16.0 23.1 2,3 17 28.3 13.8 33.2 13.9 26.7 3,4 15.3 32.6 10.4 25.0 9.3 19.9 4,5 13 30.3 10.1 25.2 7.4 18.6 5,6 11.7 29.6 9.8 25.4 4.6 15.1 (7.3)† 6,7 10.3 27 9.7 25.5 4.5 15.3 7,8 9.7 25.5 9.6 25.6 4.5 15.5 8,9 9.6 25.6 9.5 25.6 4.5 15.7 9,10 9.5 25.7 9.4 25.6 4.5 15.8 + · ··· a:ForH3O (H2O)n: n =1, , 7 data are from [29], and for n>7 from the Thomson model. Other data − · ··· from [30] and [31] are used for sensitivity analysis. b:ForNO3 (H2O)n: n =1, , 3 data are from [37], − · and for n>3 from the Thomson model. c:ForNO3 (HNO3)n: data are from [38] for n =1, from [39] for n =2, ··· , 6, and from the Thomson model for n>6. †: The entropy value in brackets is from [39], but is considered unreliable. Accordingly, the entropy predicted by the Thomson model is used for this cluster as well. mass distribution, owing to the lower amounts of nitric acid vapor in air. It is crucial to note, however, that in either case, the time required to establish the characteristic ion mass spectrum is much shorter than the ion-ion recombination lifetime, ∼ 300 s. This implies that distinct relative ion populations within cluster families will always exist in an air mass that is continuously ionized. In Figure 7, typical size distributions are displayed for clusters of nitrate core ions with water (left panel) and with nitric acid (right panel). The relative nitrate/water cluster distribution reaches a quasi-equilibrium state in roughly 50 milliseconds, after which the spectrum of the predominant clusters remains fixed. The equilibration time for nitrate/nitric acid clusters is also fairly short–of the order of seconds. While over the time scales indicated, the relative cluster distributions effectively reach a steady- state (identified by t = ∞ in the figure legend), the absolute cluster concentrations continue to build up overamuch longer period, which is determined by the mean ion lifetime (roughly 300 s for conditions relevant to the lower stratosphere). The most probable nitrate/water clusters are found to have about 6-10 ligands. This range varies with the water vapor concentration and ambient temperature. By contrast, the nitrate/nitric acid clusters typically have three ligands for the conditions tested. More water molecules are collected in part because the partial pressure of H2Oismuch greater than that of HNO3. For ion nucleation, water clusters of critical size contain hundreds to thousands of ligands (for a range of typical stratospheric water vapor concentrations and environmental conditions). The critical number is determined by the maximum in the Gibbs free energy change associated with cluster formation [26]. Figure 8 shows the free energy curves corresponding to the hydronium ion series at several specific temperatures. For temperatures above 170 K,adistinct peak occurs in ∆G0,n (although the peak lies off the chart for T>175K). The number of ligands at this local peak defines the size of the critical cluster (nucleation embryo), as indicated in the figure. Similarly, the local minimum in the free energy defines the most probably cluster size. That is, the most abundant ion clusters will lie within the free energy “well,” which is typically far removed from the critical cluster size. The difference between the 35 130 NO− (H O) p = 100 mbar NO− (HNO ) p = 100 mbar 3 2 n 3 3 n 120 [H O] = 5 ppmv T = 175 K [HNO ] = 10 ppbv T = 175 K 30 2 3 110 t=.1 [ms] t= 5 [ms] 100 25 t=.2 [ms] t=50 [ms] 90 t=.3 [ms] t=0.1 [s] 80 20 t=.4 [ms] t=0.2 [s] 70 t= 1 [ms] t=0.5 [s] 60 15 t=50[ms] t= 5 [s] t= ¥ t=¥ 50 10 40 30 Fractional Aboundances [%] Fractional Aboundances [%] 5 20 10 0 0 0 5 10 15 20 2 4 6 8 10 n, # of ligands per cluster Fig. 7: Relative concentrations (in percent) of nitrate/water clusters (left panel) and nitrate/nitric acid clusters (right panel) as a function of time, assuming the production of initially bare nitrate core ions, under the conditions indicated in the legend. From D’Auria and Turco [19]. 100 H O+ (H O) 3 2 n 190 K p = 100 mbar 50 220 K 180 K 210 K 0 [kcal/mol] 200 K 0,n G 175 K D −50 170 K −100 0 1 2 3 4 10 10 10 10 10 n, # of ligands per cluster Fig. 8: Gibbs free energy changes for the formation of hydronium ion clusters with n water ligands. At each temperature, the vertical dashed line corresponding to the local maximum in the free energy defines the size of the critical cluster; the dashed line at the local minimum identifies the size of the most probable cluster. energy maximum and energy minimum defines the energy “barrier” to nucleation. Notice that, as the temperature decreases in Figure 8, the energy barrier is systematically lowered. If temperatures were to fall below ∼170 K, the barrier would disappear and the hydronium ions would nucleate freely. In the figure, the size of the critical cluster increases dramatically as the temperature rises. At 175 K, more than 400 water ligands would be needed to build a nucleation embryo. Because these cluster sizes are so large, and the corresponding cluster concentrations are so small, the nucleation of pure water onto hydronium ions will not occur in the lower stratosphere even in the most extreme situations. Figure 9 summarizes the behavior of the three cluster families of interest for a wide range of temperatures at 100 mbar, assuming 5 ppmv of water vapor and 10 ppbv of nitric acid vapor (as in previous figures). As the temperature increases, the peak of each size distribution tends to shift toward smaller cluster sizes. This is to be expected, inasmuch as higher temperatures create conditions less favorable to condensation. It is worth noting that at 180 K, roughly 10% of the hydrated hydronium and nitrate clusters have 10 or more ligands. The nitrate/nitric acid clusters are much smaller (in terms of the number of ligands, not necessarily mass), even at very low temperatures. It is likely, however, that these clusters will take up water from the environment and grow much larger (refer, for example, to the nitrate/water series). Among other 21 10 NO− (H O) 18 3 2 n+9 20 15 10 40 30 12 70 50 20 9 H O+ (H O) 3 2 n 30 40 6 50 60 , # of ligands per cluster n 3 4 180 19010 200 210 3 80 50 80 2 60 NO− (HNO ) 20 3 3 n 1 170 180 190 200 210 220 T [K] Fig. 9: Equilibrium cluster size distributions for three common series of atmospheric cluster ions. The plot is divided into lower and upper panels, where the vertical scale (indicating the number of ligands) is linear in the upper panel and logarithmic in the lower panel. Data are displayed as isopleths corresponding to the percentage of the total number of ions in a family having a specific number of ligands, at each temperature. The cluster − · + · series are identified in the figure: NO3 (H2O)n (upper contours in the top panel); H3O (H2O)n (lower − · contours in the top panel); and NO3 (HNO3)n (contours in the lower panel). An offset of n =9ligands for the nitrate/water clusters separates them from the hydronium/water ions. From D’Auria and Turco [19]. things, the vapor pressures above solutions of HNO3 and H2O are generally much lower than the vapor pressures over the pure substances, suggesting that such mixtures –even on microscopic scales–will be more stable. Information on mixed water/nitric acid cluster ions is sparse. In one case, stratospheric mass spec- trometric data were used to infer the free energy for the (1,1) nitrate-based H2O/HNO3 cluster [46]. Larger negatively-charged mixed ion clusters have not been detected under midlatitude conditions. Like- wise, nitric acid attached to ambient hydronium clusters has not been reported. However, Castleman and co-workers have carried out studies of water/nitric acid clusters formed around hydronium core ions (e.g., [47], [48], [49]). They found that nitric acid is readily taken up by ions (of both signs) containing a sufficient number of water ligands (typically four for the first nitric acid ligand, eight for the second, and somewhat fewer water molecules per additional HNO3 ligand). These investigations suggest minimum free energies for clusters having H2O/HNO3 molar ratios approaching that of the nitric acid trihydrate ice (i.e., 3:1), although the data are currently incomplete. We are currently simulating the properties of mixed H2O/HNO3 nitrate and hydronium clusters of the type studied by Castleman et al. to extend their results. In the polar winter stratosphere, where sufficiently low temperatures are reached, newly formed hydronium and nitrate core ions may quickly become hydrated with 10 or more water molecules (Fig. 9). Some ions may collect many more. The largest water clusters would readily take up nitric acid vapor, as has been noted experimentally. At temperatures approaching 195 K, perhaps 2-3 nitric acid molecules would initially condense on the largest hydrated ions, further stabilizing these species. As temperatures dropped below 195 K, additional HNO3 and H2Owould be taken up. Even a relatively small population of charged clusters with 5-6 HNO3 molecules (and many more H2O ligands) constitutes a potentially im- portant source of freezing nuclei at temperatures in the vicinity of 190 K [18]. At the lowest temperatures attainable, it is not inconceivable that some of the largest clusters would nucleate into new aerosols. 4.2 Uncertainty analysis The sensitivity of the size distribution calculations to uncertainties in the baseline thermodynamic pa- rameters (Table 3) has been explored. For this purpose, the hydronium cluster size distributions were recalculated using an alternative set of thermodynamic data composed of measurements from Cunning- ham et al. [30] and Lau et al. [31] (refer to Table 2). These data, and those of Kebarle et al. [29] were collected over a period of 15 years by the same group using similar techniques. Nevertheless, in terms of the equivalent standard Gibbs free energy, the published data vary by roughly ±2 − 3kcal/mol (± 10%), which is representative of the differences found among other measurements (e.g., refer to Fig. 2). Ion mass distributions for the baseline and alternate cases are compared in Figure 10. The maxi- mum variation in hydronium cluster fractions is ±15%. Most of this difference arises from a subtle shift in the peak of the mass distribution by a fraction of one ligand. Hence, the degree of precision in the ther- modynamic data base appears to be adequate for the present analysis. In the case of nitrate cluster ions, insufficient data are available to reasonably evaluate uncertainty in the size distributions. However, if the thermodynamic errors are of the same order as for the hydronium family, the accuracy will be similar. Most likely, the thermodynamic parameters for nitrate clusters are not as accurate. Further, no laboratory data exist above n =6.Onthe other hand, the nitrate cluster distribution in Figure 9 is constrained to a small range of ligands, and appears to be less sensitive to thermodynamic errors for the range of con- ditions studied. Natural variability in temperatures, and water or nitric acid vapor concentrations, would produce fluctuations in the cluster distributions comparable to the variations cited above. 8 7 0 5 5 6 0 10 15 10 5 5 −5 0 −15 −10−5 10 5 −10 % difference a)−b) 5 0 −5 −5−10 4 −10 3 −5 170 175 180 185 190 195 200 205 210 215 220 8 0 30 7 20 10 50 40 30 6 40 50 20 10 20 30 40 30 10 6050 50 5 40 50 60 a) with Ref. [28] 20 30 4 10 40 20 30 3 10 170 175 180 185 190 195 200 205 210 215 220 n, # of ligands per cluster 8 10 7 30 20 10 30 6 40 40 10 20 3020 50 40 5 60 60 50 10 30 5040 b) with Ref. [29] and [30] 20 4 10 30 5040 20 3 10 170 175 180 185 190 195 200 205 210 215 220 T [K] Fig. 10: Size distributions of hydronium ions, and differences associated with uncertainties in thermodynamic parameters, as a function temperature. The numbers marking the isolines in the two lower panels (a and b) give the percentage of the total ion population having a specific number of ligands. Panels a) and b) correspond to the baseline and alternate thermodynamic models, respectively (see the text). The differences between the isopleths in panels a) and b) are shown in the uppermost panel, and represent the absolute difference in relative populations (the maximum differences being roughly ±15%. Potential errors in the predicted relative concentrations of the most abundant species are restricted because the total number of clusters is controlled by ion recombination, not cluster thermodynamics. Conversely, errors in the cluster concentrations in the “wings” of the size distribution can be much larger. For example, in the case of the hydronium series, differences by factors of 2-3 and more in absolute concentrations are found at cluster sizes well above 10 ligands. The concentrations of these very large species are extremely small, however, and they have no atmospheric role. 4.3 Polar processes Using the model described above, we can explore the role of ionic clusters in stratospheric chemistry and microphysics. Kawa and coworkers [50] have suggested that the decomposition of N2O5 on hydrated cluster ions is a source of nitric acid in the polar winter stratosphere. To test this hypothesis, the present thermodynamic/kinetic model can be applied to estimate the rates of cluster-induced heterogeneous de- composition of N2O5 at 40 km. Hamill and Turco [18] proposed that hydrated nitrate/nitric acid clusters may act as freezing nuclei when they impinge on supercooled polar stratospheric cloud (PSC) particles. − · In this case, the collision frequencies of NO3 (HNO3)n cluster ions with preexisting supercooled PSC droplets can be calculated using the model developed here. 4.3.1 Heterogeneous conversion of N2O5 The rate of reaction of N2O5 with proton hydrates may be quantified using hydronium ion mass distribu- tions corresponding to the environmental conditions encountered at 40 km (i.e, ∼ 3 mbar and ∼ 250 K). These distributions are easily derived using our hybrid ion cluster model. The results are displayed in Fig. 11 for both the baseline and alternate thermodynamic data sets utilized in previous sections. 100 100 t=0.1 [ms] t=0.1 [ms] t=0.2 [ms] 90 t=0.2 [ms] 90 a) t=0.3 [ms] b) t=0.3 [ms] t=0.4 [ms] t=0.4 [ms] 80 80 t=1 [ms] t=1 [ms] t=50 [ms] t=50 [ms] 70 t= 70 ¥ t=¥ 60 60 C(n,t)) [%] 50 50 C(n,t)) [%] N n=0 N n=0 S S 40 40 (C(n,t)/ 30 30 (C(n,t)/ 20 20 10 10 0 0 0 1 2 3 4 5 6 0 1 2 3 4 5 6 7 n, # of ligands per cluster n, # of ligands per cluster Fig. 11: Stratospheric hydronium water cluster size distributions. Panel a): relative concentrations correspond- ing to the baseline thermodynamic data in Table 3 ([29]); Panel b): concentrations corresponding to the ther- modynamic data from [30] and [31]. In both cases, ambient conditions at an altitude of ∼ 40 km are assumed. Rate coefficients for the reaction of N2O5 with hydronium-based ions have been reported by Bohringer¨ and coworkers [51] for the first 6 hydrated species. Because the cluster concentrations for n>6 are negligible (Fig. 11), rate coefficients for these massive ions are not required. However, only upper-limit rate constants are specified for ions having 3 or more water ligands. It follows that the derived loss rates for the decomposition of N2O5 on proton hydrates will represent upper limits for this process. Combining the modeled cluster ion size distribution with the known reaction rate coefficients, the −8 −1 overall loss rate for N2O5 is found to be less than 3 × 10 s . This upper limit applies to both of the cluster size distributions in Figure 11. The fastest reactions between N2O5 and the proton hydrates occur for the smallest clusters, with n<3.Onthe other hand, PH clusters in the middle stratosphere are dominated by hydronium with three water ligands. Consequently, the overall reaction efficiency of N2O5 with such clusters is quite low–in fact, two orders of magnitude too low to influence the behavior of odd-nitrogen in the stratosphere. 4.3.2 Ion cluster interactions with PSCs Hamill and Turco [18] recently suggested that nitrate ion clusters having a composition similar to that of the nitric acid trihydrate may act as freezing nuclei for type 1b PSC particles consisting of supercooled − · nitric acid aqueous solutions. Here, we can estimate the collision frequencies of NO3 (HNO3)n clus- ters with a typical type 1b aerosol (i.e.,adroplet of about 1 µm diameter). Considering the polar winter environment, temperatures close to 190 K are expected at 100 mbar pressure. The corresponding ni- trate/nitric acid cluster ion distribution is determined by the thermodynamic data in Table 3 and the solutions to Eqs. 2. The rates of encounter of a PSC type 1b aerosol with thermally driven clusters of each size are then readily calculated. The collision frequencies for clusters with 0 to 5 nitric acid ligands turn out to be, respectively, 5 × 10−5 s−1, 4 × 10−5 s−1, 2 × 10−2 s−1, 3 × 10−1 s−1, 6 × 10−4 s−1, and 2 × 10−8 s−1. These estimates suggest that supercooled PSC droplets will encounter a charged embryo contain- ing 5-6 nitric acid molecules (including the nitrate core ion, and most likely a substantial number of water molecules) on time scales of hours to days. This theoretical range of collision frequencies is consistent with the rates of droplet freezing derived from observations by Tabazadeh et al. [52]. It follows that large nitrate cluster ions may have a role in the observed phase transitions of polar stratospheric clouds. 5. CONCLUSIONS We have developed a hybrid kinetic/thermodynamic model to investigate the behavior of atmospherically- prominent families of large ion-molecule clusters. The methodology is generally applicable to charged molecular aggregates of all kinds. Such clusters represent a fundamental state of matter, associated with an initial phase transition from vapors to particles. Moreover, charged aggregates play an important role in atmospheric chemistry and microphysics. In our approach, laboratory thermodynamic measurements and quantum mechanical structural cal- culations are combined with a phenomenological representation of charged macroclusters–the Thomson model–to define the thermodynamic properties of ionic species over the entire size range from molecular ions to nanoparticles. We show how this hybrid data base can be used to constrain the kinetic equations describing the evolution of cluster populations for a wide range of environmental states. We focus on three key ion families; hydronium/water, nitrate/water, and nitrate/nitric acid. Size distributions for these families are obtained under polar stratospheric conditions for the first time, and are employed to assess possible effects on upper-atmospheric composition. The analysis in this paper demonstrates that a practical and accurate description of naturally- occurring ionic clusters can be constructed by combining three sources of information: small cluster ther- modynamics based on laboratory measurements, intermediate cluster energetics derived through quan- tum mechanical simulations, and large cluster phenomenology defined by macroscopic physical/chemical properties. In particular, we find that quantum mechanical simulations provide an extremely useful tool for investigating charged clusters when reliable laboratory data are unavailable. In some instances, the simulations can be partially validated within a cluster family using measurements of the smallest species. An important limitation to the application of quantum mechanical techniques is the heavy computational burden required to simulate large, complex molecular clusters. For the ion series of interest, we show that the characteristic size distributions are not highly sen- sitive to existing uncertainties in the thermodynamic model, at least for the most abundant clusters in a sequence. Greater uncertainty is associated with very large species, although the corresponding concen- trations are extremely low. We quantify the behavior of the ion population corresponding to variations in ambient temperature and pressure. At temperatures reached in the upper atmosphere, a fraction of the background water and nitric acid clusters can achieve massive sizes. We conclude, based on newly pre- dicted ion-hydrate size distributions, that certain chemical reactions–in particular the reaction of N2O5 with hydronium ions–will not significantly perturb the nitrogen cycle in the middle and upper strato- sphere. However, we also find that large nitrate clusters composed of water and nitric acid may reach high enough concentrations to affect the properties of polar stratospheric clouds, which in turn control stratospheric ozone loss in the winter polar stratosphere. The methodology proposed here can be extended to any ion cluster series for which sufficient thermodynamic data are available, or can be generated, including complex clusters composed of mixed ligands. Indeed, the data base for such species has been growing dramatically owing to the impor- tance of atmospheric aerosols in cloud microphysics and climate change, which has motivated numerous laboratory and field experiments, as well as theoretical studies of aerosol nucleation. The hybrid ap- proach delineated in this work is general enough to treat charged clusters throughout the troposphere and stratosphere. Examples in the text demonstrate that a variety of practical problems can be investigated with such a model. Additional research is needed to support this approach, however, including in situ characterization of ion mass spectra and compositions, laboratory investigations of basic cluster ther- modynamics and kinetics, and detailed simulations of intermediate-sized charged cluster structures and energetics via ab initio and Monte Carlo techniques. ACKNOWLEDGMENTS This work has been funded by NASA under grant NAG1-1899, and the NSF under grant ATM-00-70847. RD is also supported by NASA Earth System Science Fellowship ESS/00-0000-0080. The authors also acknowledge Dr. Kendall Houk for critical advice on and access to quantum mechanical solutions for the structures of ionic clusters. References [1] Thomson, J. J., Conduction of Electricity through Gases, Cambridge University press, Cambridge, 1906. [2] Ferguson, E. E., F. C. Fehsenfeld and D. L. Albitron, Ion chemistry of the Earth’s atmosphere, Gas Phase Ion Chemistry, Volume 1, edited by M. T. Bowers, pp. 45-83, Academic Press, Inc., London, 1979. + · [3] Dotan, I., J. A. Davidson, F. C. Fehsenfeld and D. L. Albritton, Reactions of O2 O2 with + · CO2, O3, and CH4 and O2 O3 with H2O and CH4 and their role in stratospheric ion chemistry, J. Geophys. Res., 83, 4063-4068, 1978. [4] Arijs, E., D. Nevejans and J. Ingels, Unambiguous mass determination of major stratospheric posi- tive ions, Nature, 288, 684, 1980. [5] Eisele, F. L., First tandem mass spectrometric measurement of tropospheric ions, J. Geophys. Res., 93, 716-724, 1988. [6] Tanner, D. J., and F. L. Eisele, Ions in oceanic and continental air masses, J. Geophys. Res., 96, 1023-1031, 1991. [7] Arnold, F., and G. Henschen, First mass analysis of stratospheric negative ions, Nature, 257, 521, 1978. [8] Arnold, F., A. A. Viggiano and H. Schlager, Implications for trace gases and aerosols of large negative ion clusters in the stratosphere, Nature, 297, 371-376, 1982. [9] Fehsenfeld, F. C., and E. E. Ferguson, Thermal energy positive ion reactions in a wet atmosphere containing ammonia, J. Chem. Phys., 59, 6272-6276, 1973. [10] Fehsenfeld, F. C., C. J. Howard and A. L. Schmeltekopf, Gas phase ion chemistry of HNO3, J. Chem. Phys., 63, 2835-2841, 1975. − − [11] Davidson, J. A., F. C. Fehsenfeld and C. J. Howard, The heats of formation of NO3 and NO3 association complexes with HNO3 and HBr, Int. J. Chem. Kinet., 9, 17-29, 1977. [12] Brasseur, G., and A. Chatel, Modelling of stratospheric ions: A first attempt, Ann. Geophys., 1, 149-160, 1983. [13] Kawamoto, H., and T. Ogawa, A steady state of negative ions in the lower stratosphere, Planet. Space Sci., 32, 1223-1233, 1984. [14] Kawamoto, H., and T. Ogawa, First model of negative ion composition in the troposphere, Planet. Space Sci., 34, 1229-1239, 1986. [15] Beig, G., S. Walters and G. Brasseur, A two-dimensional model of ion composition in the stratosphere-1. Positive ions, J. Geophys. Res., 98, 12,767-12,773, 1993. [16] Beig, G., and G. Brasseur, Model of tropospheric ion composition: A first attempt, J. Geophys. Res., 105, 22,671-22,684, 2000. + · [17] Viggiano, A. A., F. Dale and J. F. Paulson, Proton transfer reactions of H (H2O)n=2,11 with methanol, ammonia, pyridine, acetonitrile, and acetone, J. Chem. Phys., 88, 2469-2477, 1987. [18] Hamill P., and R. P. Turco, Ion nucleation of NAT in ternary system polar stratospheric clouds, Trans. Amer. Geophys. Union, 81 (#19, May 9, 2000) Suppl., S97, 2000. [19] D’Auria, R., and R. P. Turco, Ionic clusters in the polar winter stratosphere, Geophys. Res. Lett.,in press, 2001. [20] Zhao, J., O. B. Toon and R. P. Turco, Origin of condensation nuclei in the springtime polar strato- sphere, J. Geophys. Res., 100, 5215-5227, 1995. [21] Bates, D. R., Recombination of small ions in the troposphere and lower stratosphere, Planet. Space Sci., 30, 1275-1282, 1982. [22] Cole, R. K., and E. T. Pierce, Electrification in the Earth’s atmosphere for altitudes between 0 and 100 kilometers, J. Geophys. Res., 70, 2735-2749, 1965. [23] Paltridge, G. W., Experimental measurements of the small-ion density and electrical conductivity of the stratosphere, J. Geophys. Res., 70, 2751,2761, 1965. [24] Paltridge, G. W., Stratospheric small-ion density measurements from a high-altitude jet aircraft, J. Geophys. Res., 71, 1945-1952, 1966. [25] Ferguson, E. E., Laboratory measurements of ionospheric ion-molecule reaction rates, Rev. Geo- phys., 12, 703-713, 1974. [26] Castleman, A. W., Jr., and I. N. Tang, Role of small clusters in nucleation about ions, J. Chem. Phys., 57, 3629-3638, 1972. [27] Castleman, A. W., Jr., P. M. Holland and R. G. Keesee, The properties of ion clusters and their relationship to heteromolecular nucleation, J. Chem. Phys., 68, 1760-1767, 1978. [28] Holland, P. M., and A. W. Castleman, Jr., Thomson equation revisited in light of ion-clustering experiments, J. Phys. Chem., 86, 4181-4188, 1982. [29] Kebarle, P., S. K. Searles, A. Zolla, J. Scarborough and M. Arshadi, The Solvation of the hydrogen ion by water molecules in the gas phase. Heats and entropies of solvation of the individual reactions: + + H (H2O)n−1 +H2O=H (H2O)n, J. Am. Chem. Soc., 89, 6393-6399, 1967. + [30] Cunningham, A. J., J. D. Payzant and P. Kebarle, A kinetic study of the proton hydrate H (H2O)n equilibria in the gas phase, J. Am. Chem. Soc., 94, 7627-7632, 1972. [31] Lau, Y. K., S. Ikuta and P. Kebarle, Thermodynamics and kinetics of the gas-phase reactions: + + H3O (H2O)n−1 +H2O=H3O (H2O)n, J. Am. Chem. Soc., 104, 1462-1469, 1982. [32] Meot-Ner M., and C. V. Speller, The filling of solvents shells in cluster ions. 1. Thermochemical criteria and the effects of isomeric clusters, J. Phys. Chem., 90, 6616-6624, 1986. [33] Shi, Z., J. V. Ford, S. Wei and A. W. Castleman, Jr., Water clusters: Contributions of binding energy and entropy to stability, J. Chem. Phys., 99, 8009-8015, 1993. [34] Klots, C. E., Kinetic methods for quantifying magic (atomic clusters), Z. Phys. D 21, 335-342, 1991. + ≤ [35] Yang, X., and A. W. Castleman, Jr., Large protonated water clusters H (H2O)n (1 n>60): The production and reactivity of clathrate-like structures under thermal conditions, J. Am. Chem. Soc., 111, 6845-6846, 1989. [36] Magnera, T. F., D. E. David and J. Michl, The first twenty-eight gas-phase proton hydration ener- gies, Chem. Phys. Lett, 182, 363-370, 1991. [37] Lee, N., R. G. Keesee and A. W. Castleman, Jr., The properties of clusters in the gas phase. IV. − Complexes of H2O and HNOx clustering on NOx , J. Chem. Phys., 72, 1089-1094, 1980. − − [38] Davidson, J. A., F. C. Fehsenfeld and C. J. Howard, The heats of formation of NO3 and NO3 association complexes with HNO3 and HBr, Int. J. Chem. Kinet., 9, 17-29, 1977. − · ≤ [39] Wlodek, S., Z. Luczynski and H. Wincel, Stabilities of gas-phase NO3 (HNO3)n, n 6, clusters, Int. J. Mass Spectrom. Ion Phys., 35, 39-46,1980. [40] Arnold, F., and S. Qiu, Upper stratosphere negative ion composition measurements and infrared trace gas abundances, Planet. Space Sci., 32, 169-177, 1984. [41] Kohn, W., and L. J. Sham, Self-consistent equations including exchange and correlation effects, Phys. Rev., 140A, 1133, 1965. [42] Becke, A. D., A new mixing of Hartree-Fock and local density-functional theories, J. Chem. Phys., 98, 1372-1377, 1993. [43] Frisch, M. J., G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, V. G. Zakrzewski, J. A. Montgomery, R. E. Stratmann, J. C. Burant, S. Dapprich, J. M. Millam, A. D. Daniels, K. N. Kudin, M. C. Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi, R. Cammi, B. Mennucci, C. Pomelli, C. Adamo, S. Clifford, J. Ochterski, G. A. Petersson, P. Y. Ayala, Q. Cui, K. Morokuma, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V. Ortiz, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, C. Gonzalez, M. Challacombe, P. M. W. Gill, B. G. Johnson, W. Chen, M. W. Wong, J. L. Andres, M. Head-Gordon, E. S. Replogle and J. A. Pople, Gaussian 98 (Revision A.7), Gaussian, Inc., Pittsburgh PA, 1998. + [44] Newton, M. D., Ab initio studies of the hydrated H3O ion. II. The energetics of proton motion in higher hydrates (n=3-5), J. Chem. Phys., 67, 5535-5544, 1977. [45] Wei, D., and D. R. Salahub, Hydrated proton clusters and solvent effects on the proton transfer barrier: A density functional study, J. Chem. Phys., 101, 7633-7642, 1994. [46] Viggiano, A. A., and F. Arnold, The first height measurement of the negative ion composition of the stratosphere, Planet. Space Sci., 29, 895, 1981. [47] Castleman, A. W., Jr., Experimental studies of ion clustering: Relationship to aerosol formation processes and some atmospheric implications, J. Aerosol Sci., 13, 73-85, 1982. [48] Castleman, A. W., Jr., P. M. Holland and R. G. Keesee, Ion association processes and ion clustering: Elucidating transitions from the gaseous to the condensed phase, Radiat. Phys. Chem., 20, 57-74, 1982. [49] Gilligan, J. J., and A. W. Castleman, Jr., Acid dissolution by aqueous surfaces and ice: insights from a study of water cluster ions, J. Phys. Chem., 105, 5601-5605, 2001. [50] Kawa, S. R., J. B. Kumer, A. R. Douglass, A. E. Roche, S. E. Smith, F. W. Taylor and D. J. Allen, Missing chemistry of reactive nitrogen in the upper stratospheric polar winter, Geophys. Res. Lett., 22, 2629-2632, 1995. [51] Bohringer¨ H., D. W. Fahey, F. C. Fehsenfeld and E. E. Ferguson, The role of ion-molecule reactions in the conversion of N2O5 to HNO3 in the stratosphere, Planet. Space Sci., 31,185-191, 1983. [52] Tabazadeh, A., E. J. Jensen, O. B. Toon, K. Drdla and M. R. Schoeberl, Role of the stratospheric polar freezing belt in denitrification, Science, 291, 2591-2594, 2001. THE PRODUCTION OF ATMOSPHERIC NITRIC OXIDE BY COSMIC RAYS & SOLAR ENERGETIC PARTICLES Barry J. Kellett Space Science & Technology Department, Rutherford Appleton Laboratory, Chilton, Didcot, Oxon., UK Abstract Galactic Cosmic Rays (GCRs) deposit their energy throughout the atmosphere but peaking in the low Stratosphere and upper Troposphere. GCR ionisation leads to the production of nitric oxide (NO) at significant levels which are also modulated in anti-phase with the solar cycle. This accounts for approximately half the NO at high latitudes with a large ∼ 11-year modulation. Solar Energetic Particles (SEPs) Events occur sporadically but are more fre- quent around solar maximum. They interact with the atmosphere in the 30-60 km range but occasionally they can penetrate down to below 20 km. Because they show a very dramatic onset with huge increases in energetic protons, SEPs are useful for studying the effects of energetic particles on atmospheric chem- istry. Lightning is also a potentially important source of NO and is also possi- bly correlated with GCRs. Ice core data shows that SEP generated nitrates can reach the ground/low atmosphere in large quantities. 1. INTRODUCTION This is intended to be a short overview of how cosmic rays do affect atmospheric chemistry and specifi- cally with regard to the various oxides of nitrogen. To illustrate this in a more direct manner, I will focus on “Solar Energetic Particle” (SEP) Events1. These are generally rather less energetic than “true” cosmic rays, but they display much greater dynamic variability that allows us to follow some of their effects on the atmosphere. I will also touch on the role of OH and on general ozone effects and conclude with a look back in time by using some recent ice core data, going 400 years into the Sun’s active past. Probably the first person to discuss the possible climatological effects of cosmic rays was Edward Ney [1]. Ney stressed that because the solar modulation of cosmic rays affected low energy particles more than higher energy ones, the atmospheric change in ionisation would show a latitude effect. Having demonstrated that cosmic rays were indeed modulated at all latitudes, he went on to speculate further using a diagram (figure 1). He wondered if there was a connection between ionisation and thunderstorm activity (for example). If so, then a solar cycle modulation might be detectable in the climate data. It should perhaps be stressed here that Ney’s paper was written in 1959. Ney commented that his diagram was only a “suggestion” and he was confident that climatologists should be able to come up with many such scenarios and that these could then be tested against the cosmic ray and weather data. Ney concluded by noting that the meteorological variable subject to the largest solar cycle modulation in the denser layers of the atmosphere (i.e. greater than 1 mb pressure) is the ionisation produced by cosmic rays and that it should be worthwhile investigating the possible effects of changes in this variable on the climate. In this paper, I would like to review the role of cosmic rays (and solar energetic particles) in the production of oxides of nitrogen and to then suggest my own “diagrammatic scenario”, building on the foundations laid by Edward Ney in 1959. In particular, I would like to stress the role that lightning plays in generating nitric acid in the troposphere and whether this could be influenced by the modulation in atmospheric ionisation and therefore coupled to the solar cycle modulation (via the crucial role played by cosmic rays). 1Solar energetic particles events are also sometimes referred to as Solar Proton Events Ð SPEs. Fig. 1: Ney’s diagrammatic scenario illustrating one possible way that solar activity might be coupled to the climate record. The first two links (shown by solid arrows) Ney believed were already firmly established (this was in 1959). However, the last three links were more speculative (as indicated by the shaded arrows and question marks). 2. SOLAR ENERGETIC PARTICLE EVENTS Solar energetic particle (SEP) events are, in these modern times, inextricably linked with the term “space weather”. With the construction of the International Space Station (ISS) and a permanent manned pres- ence in space planned for the near future, it is vital that solar scientists are able to give advanced warning of any solar activity that might be potentially harmful to astronauts and scientists living on the ISS. In some cases (as I will try and show below), the warning given might be very little indeed. In the past, perhaps the best (and certainly the most beautiful and visual), demonstration that the Sun was “up to something” was the aurora or Northern Lights. These are perfectly harmless events that indicate energetic electrons are streaming into the atmosphere (mostly at the poles but occasionally at more accessible mid-latitudes) and causing the nitrogen and oxygen molecules in the air to fluoresce and emit beautiful ribbons and curtains of flickering coloured light. In terms of space weather forecasting, the SOHO spacecraft is currently in the front line with its vantage point some 1.5 million km closer to the Sun than Earth (a mere 1% closer). As an example of how dramatic and rapid SEP events can be, lets look at an event from Bastille Day (July 14th). Bastille Day is a national holiday in France and is typically celebrated with parades and parties and firework displays. In the year 2000 (the 210th anniversary), the Sun arranged its own special “fireworks” display. Alarge flare was seen to erupt by SOHO to start at 10:12 UT and it peaked at 10:24. It was a3Bflare in the optical classification scheme and an X5 flare in the X-ray band - both of which are fairly impressive events (it is perhaps worth noting that July 2000 was fairly close to the Sun’s maximum phase in its 11-year solar sunspot and activity cycle). The flare was from NOAA active region 9077, which was very close to the centreline of the Sun (as viewed from Earth) and just north of the Sun’s equator). Solar flares are very often associated with coronal mass ejections (CMEs). CMEs are relatively cool material and once released from the Sun rapidly expand to become “clouds” travelling through interplanetary space (along with the Sun’s normal plasma emission - the solar wind - which typically travels at a velocity of about 450 km/s). However, CMEs associated with large flares are very often “fast” events - that is they have velocities in excess of 7-800 km/s. When these CMEs start propagating through the normal solar wind, they start to “sweep up” material and compress/stretch the imbedded solar/interplanetary magnetic field. This leads to shocks forming and the front of the CME becomes asite of in situ particle acceleration. Therefore, such a fast CME produces a large surge of energetic particles (mostly protons) - hence, a solar energetic particle (or solar proton) event. Fig. 2: Images of the SOHO LASCO coronagraph of the halo CME associated with the Bastille Day flare. Notice how the images are very quickly covered with a “snow storm” - i.e. direct impacts on the CCD camera of the energetic particles produced by the CME. The images were taken at (top) 09:18, 10:18, 10:42, 11:18, (bottom) 12:47, 21:10, and 22:57. The bright white ring in the 11:18 image is the CME coming directly towards the Earth that is also visible in the subsequent images. The size of the Sun (obscured) in these images is shown by the small white circle at the centre of each frame. From the ground, if we want to see the solar corona (i.e. the Sun’s hot outer tenuous atmosphere), we normally have to wait for a solar eclipse. However, from space (and indeed high mountains) it is possible to create artificial eclipses by obscuring the Sun’s bright visible disk. Such an instrument is called a coronagraph. The coronagraph imager on SOHO is called LASCO and it detected a halo CME from the Bastille Day flare at about 10:54 (figure 2). A halo CME is a CME that the Sun launches essentially directly towards the Earth and hence quickly produces a halo around the Sun (see the top right image of figure 2). However, almost immediately after LASCO detected the CME it was quickly “blinded” by the energetic particles from the event itself - the SEPs (clearly visible as the “snow storm” in the last four images). The SEP event itself is shown in figure 3 along with the CME shock/disturbance in the solar wind that took rather longer to reach the Earth. The SEP particles reached SOHO in around 40 minutes and went on to the reach the Earth in a few minutes later. The steep rise in this event shows an almost instantaneous increase of around ×50,000 in all energy bands and was the biggest SEP event since the previous solar maximum in 1991. If you were an astronaut on the ISS “outside” doing some work when the flare erupted, you would have had very little warning of the SEP event and the huge dose of radiation coming towards you. Remember, light takes about 8 minutes to get from the Sun to Earth/SOHO — the data then had to be collected and stored on board SOHO before it could be relayed back to Earth, analysed in “realtime” and a possible warning given — all this would have about 30 minutes. This would then give you (the astronaut) just 10 minutes to get back into the habitation module and “hide” behind the protective shielding. This emphasises that flares are not predictable and the size of a flare is similarly not predictable in advance. Having said that, active region 9077 was a large sunspot group, and a big flare was expected from it. Fig. 3: The left panel shows the almost instantaneous increases in energetic particles and electrons while the right panel detects the arrival of the CME shock at SOHO a mere 28 hours after the flare erupted. It should be noted that the event lasted many days if judged in terms of energetic particles but was measured in hours in terms of X-rays from the flare itself. This emphasises that it is the CME and not the flare that produces the bulk of the energetic particles. 3. COSMIC RAY INTERACTIONS IN THE ATMOSPHERE Galactic cosmic rays (GCRs) are energetic charged particles that originate throughout the Milky Way galaxy and for the energies that we will be interested in here GCRs are produced in supernova explosions and the subsequent supernova remnant expansion into the surrounding interstellar medium leading to shock acceleration, etc. (In fact, the acceleration mechanism is probably very similar to that at a CME shock front in an SEP event, although on a rather more dramatic scale). GCRs are about 88% protons, 10% helium, around 1% heavier ions and less than about 1% electrons. The energy flux reaching the Earth (∼10−9 W/cm2)isalmost completely negligible, but GCRs are very definitely important in their contribution to atmospheric ionisation. In particular, GCRs are able to penetrate much deeper into the atmosphere than solar ionising radiation and are the dominant ionisation process below about 60km (with an additional contribution from SEP events as we will see shortly). [NB: This statement is true down to about 3-4 km where surface radioactivity also becomes important]. For GCR energies up to about 20 GeV, the solar wind and solar heliosphere (the region of interstellar space dominated by the Sun, extending out to about 100 AU [1 AU being the average Sun-Earth distance = 150 million km]) does scatter and deflect incoming particles. Since the heliosphere responds to the ∼11 year solar sunspot activity cycle, then so does the flux of these lower energy GCRs (the GCR energy spectrum extends to beyond 1011 GeV so 20 GeV is “low” for a GCR!). This then opens up the possibility of GCRs coupling to solar activity and producing a measurable climate variation as already noted above by Ney[1]. It is perhaps worth noting that the GCR flux goes down at solar maximum and peaks at solar minimum - that is, the GCR flux is in anti-phase with the sunspot cycle. The Earth’s own magnetic field also provides an addition level of protection from these lower energy GCRs, leading to the already noted latitudinal variation in the cosmic ray data. The largest modulations are seen at high geomagnetic latitudes and the smallest modulations (i.e. sunspot maximum−→ sunspot minimum) are seen around the geomagnetic equator in regions of high “rigidity”. This magnetic rigidity means that at the highest level only GCRs with energies around 14 GeV can penetrate to the lower levels of the atmosphere/ground level. However, since this is still less than the 20 GeV energy that is solar modulated, even equatorial regions experience a modulated cosmic ray flux. 3.1 GCR IMPACT ON ATMOSPHERIC CHEMISTRY When a primary cosmic ray of energy around 1 GeV enters the atmosphere it initiates a huge avalanche of secondary particles (normally referred to as an air shower) with more than a million secondary particles produced. This flux of secondary particles increases as we traverse down through the atmosphere until we reach around 15Ð25 km (depending on magnetic rigidity and solar cycle phase) below which the flux decreases again. (The ionisation maximum height is more correctly a measure of the total atmospheric column density traversed by the secondaries, which is equivalent to saying the atmospheric pressure). In this paper it is not the ionisation itself that I wish to consider, but the atmospheric chemistry that might result from cosmic rays hitting the atmosphere. Given that the atmosphere is essentially made of nitrogen (N2), oxygen (O2) and a trace of water vapour (H2O), it shouldn’t be to surprising that the major chemical species generated by GCRs are oxides of nitrogen and hydrogen — so called NOx and HOx. [NOx —N,NO, NO2;HOx —H,OH, HO2]. There is also a secondary effect on ozone which will be briefly discussed. Warneck[2] was the first to note that GCRs would be a significant source of NOx (and particularly NO). (This was around the time that ozone destruction was beginning to be studied and aircraft emissions and other manmade contributions had already been considered as a possible culprit). However, it was Nicolet[3] who first calculated production rates of nitric oxide (NO) by GCRs and also the variation caused by the solar modulation. These calculations showed a large production and modulation effect at high geomagnetic latitude (above 50◦ latitude) at an altitude of 20 km. Some of the more important reactions are shown below. + + N2 → N+N O2 → O+O (dissociative ionisation) (1) + → + + → + + → + N +O2 O2 +N N +O2 O+NO N2 +O2 NO + NO (2) NO+O3 → O2 +NO2 NO2 +O3 → O2 +NO3 OH+NO2 + M → HNO3 + M (3) OH+O→ O2 +H H+O3 → OH+O2 OH+O3 → H2O+O2 (4) As can be seen in (3) and (4) HOx and NOx are implicated in ozone destruction. However, it is not just galactic cosmic rays that promote these atmospheric reactions — solar energetic particles can also be important. The only key difference is that SEPs produce their effects in the middle atmosphere while GCRs produce effects, in particular, in the lower stratosphere and upper troposphere. At least nine separate SEP events have been observed to produce ozone depletions in the past three solar cycles, and one of the most dramatic events was the August 1972 SEP. These ozone depletions are believed to be pri- marily due to the newly created HOx species, although the role of NOx cannot be underestimated. There is another difference between GCR and SEPs — SEPs occur most frequently around solar maximum, in particular on the rise just prior to maximum and on the fall a year or two after maximum, while GCRs peak around solar minimum. Jackman et al. [4] looked in detail at the various sources of nitric oxide and where in the atmo- sphere each process is effective (figure 4). They showed that stratospheric NO is mainly produced from the dissociation of nitrous oxide (N2O) - which is a by-product of the biological nitrogen cycle. This provides a large and relatively constant background source. However, GCRs and SEPs provide a signif- icant and variable component of NO in the middle atmosphere. In fact, at geomagnetic latitudes greater than 50◦, the GCR contribution shows a solar cycle modulation of some 50% and is also responsible for about 50% of the total NO production in the stratosphere and mesosphere. Jackman et al. [4] also made detailed calculations for some SEP events. In particular, they showed that the large event in August, 1972 (around three years after solar maximum) produced effects that extended down to 10 km. It also took the atmosphere 1 year to return to pre-event levels. 4. LIGHTNING AS A SOURCE OF NOx As can be seen in figure 4, lightning is listed as a major contributor to NO production in the atmosphere. The role of lightning was studied in more detail by Legrand et al. [5]. Lightning generates NO by the thermal decomposition of nitrogen and oxygen, and this is likely to be especially important in the Tropics and in the lower regions of the atmosphere (as can be seen in figure 4, where lightning is the only source Fig. 4: Atmospheric sources of odd nitrogen (principally NO) (from Jackman et al. [4]). Only the 0-50 km altitude range is shown here since the processes that operate at high altitudes are not relevant to the present discussion. given below 4 km). Legrand et al. [5] used a 2-dimensional model to estimate that at the tropopause, lightning contributed 30% of the total NO production at the poles, and this rose to 60% at the equator (compared to GCRs contribution of 10% or less). Charles Jackman [6] in a review talk at the Spring AGUin2001 suggested that lightning produced NO could contribute up to 1000 kilotons per year to the middle atmosphere (stratosphere + mesosphere). This would make it the single largest source of NO, exceeding the 800 kT/yr from nitrous oxide dissociation. However, there are still significant uncertainties in the exact contribution to NO production made by lightning. Crucially, galactic cosmic rays do undoubtedly play a role in lightning. As we have already seen, GCRs ionise stratospheric and tropospheric air producing free electrons and light ions. These in turn will determine the electrical conductivity of the air. It is this conductivity that allows a current to flow in the atmosphere in what is generally referred to as the global electric circuit. 4.1 THE EARTH’S GLOBAL ELECTRIC CIRCUIT The “classical” picture of the Earth’s global electric circuit is that the very high conductivity of the ionosphere (maintained by the ionising X-ray and UV radiation from the Sun) is weakly conducted back to the ground through the “fair weather” electric field. Since the ground/oceans are also good conductors, the circuit needs an “up” component to complete it. Since 1916 this upward component/generator as been assumed to be the combined global summation of all the active thunderstorms (Wilson [7]). This situation is shown schematically in figure 5 and an equivalent (simplified) electrical circuit is also shown (Makino &Ogawa [8]). It is the increased conductivity provided by GCRs that allows the circuit to operate and also allows for global redistributions to take place, since GCRs are more influential at the polar regions, this redistribution transfers the effects of GCRs to the middle and low latitudes. The global electric circuit and its links to GCRs and atmospheric conductivity have been suggested as a mechanism for increasing cloudiness (in a way that is very reminiscent of Ney’s (1959) diagrammatic scenario [shown in figure 1]). The idea was proposed by Tinsley [9], and relates to the fair weather Fig. 5: Schematic of the Earth’s electric circuit and the simplified equivalent circuit. The simplified circuit is from Makino &Ogawa [8], where r is the global Earth-Ionosphere resistance, R1 the resistance between the +ve thundercloud top to the ionosphere, R2 the resistance across the cloud and R3 is the cloud to Earth resistance. Thunderstorms can be seen to be the global electric circuit generators. currents shown in figure 5. These currents flow because of the ∼250 kV potential difference set up between the ionosphere and the ground. This current depends critically on the atmospheric conductivity of the middle and lower atmosphere (and hence on the GCR flux). Tinsley [9] suggested that when this current encounters a cloud, the current flow will see an increased electrical resistance due to the cloud (called R2 in figure 5), leading to the formation of a positive space charge on the upper surface of the cloud. It is this electrical charging that Tinsley suggests will lead to the scavenging of evaporated aerosol particles and in the freezing of droplets. This electrofreezing process can considerably enhance the overall production of ice nuclei in clouds. 4.2 LIGHTNING MODULATED BY GCRs OR THE SOLAR CYCLE? As we have already discussed, lightning is a significant source of NOx in the free troposphere. In the model of Legrand et al. [5], lightning was confined to the tropical areas (30◦N—30◦S) and also to the ground—15 km altitude range (with a relative maximum at 10 km). A global production rate of 2.8 million tonnes/year was assumed (larger than the ∼1000 kT/yr figure given above by Jackman [6], but Jackman was only quoting the amount of NO transported upward into the middle atmosphere). The key question which wasn’t addressed by the model is whether the lightning rate (and hence the resultant NOx) is modulated by solar activity and/or galactic cosmic rays. Above, we certainly suggested that GCRS should have an effect on the lightning rate since GCRs will clearly have an effect on the conductivity of the atmosphere. But is there any observational evidence for a link/modulation in the lightning data? The main “problem” in answering this question is the relative lack of a “global” lightning moni- toring network to collect the raw data needed to answer the question. Areas of the world do have some monitoring systems and these lend some support to the idea of a link between cosmic rays/solar activ- ity/solar wind and lightning frequency, but in a somewhat difficult to interpret manor. Lethbridge [10] used data from a lightning network covering much of the continental United States in a superposed epoch analyse. She found that there was a significant increase in thunderstorm activity threeÐfour days after the cosmic ray maximum (taking data on a month-by-month basis from 1956Ð1976 with seasonal trends removed). A similar correlation was also found with solar-wind magnetic sector boundary crossings and minima in monthly Kp indices. Since the sector boundary crossing also correlated strongly with cosmic ray flux, Lethbridge [10] concludes that the effect was more likely to be due to cosmic rays. However, strong solar cycle modulations have been found in parameters that relate to the global electric circuit. Measurements of the air-earth current (in fair weather and mostly taken over Lake Su- perior) in the period 1966Ð1982 show a clear solar cycle modulation with an amplitude of at least 50% (Olson [11]). The variation was in the sense that the maximum air current density was around solar mini- mum (1977), while the minimum value was seen in 1969 at the previous solar maximum. Muhleisen¬ [12] also found a variation in the ionospheric potential that was in anti-phase with the solar cycle in 11-years of balloon radiosonde measurements. His results indicated an average ionospheric potential of around 350 kV around solar minimum, falling to ∼250 kV close to solar maximum. Both these results would thus appear to be responding to the galactic cosmic ray flux and certainly the air current density result is clearly in agreement with the picture previous presented of atmospheric ionisation and its likely effects on the global electric circuit. The ionospheric potential measurements would also support a link with lightning frequency, since as previously stated lightning/thunderstorms are the generator of the global electric circuit (figure 5). So, it would seem that the answer to our question is probably “yes”, there is a plausible and probable modulation of lightning/thunderstorm activity that responds to changes in galactic cosmic ray flux. 5. SOLAR ENERGETIC PARTICLES AND ICE CORES As can be seen from the limited set of atmospheric chemical reactions shown in (1)—(4), nitric acid (HNO3)isone of the stable end points of GCR initiated processes. When nitric acid dissolves in water, − it forms nitrate ions (NO3 ). These nitrate ions can find their way into rain and snow that effectively removes them from the atmosphere. Therefore, places that are permanently frozen can preserve a time record of the rate of atmospheric nitrate production. Zeller et al. [13] have analysed nitrate ions in a snow sequence from the Ross Ice Shelf (Antarctica) dating back to 1971. The data had a resolution of 2-3 months and shows a clear annual cycle with sharp peaks in summer and broad minima in winter. The effect is believed to be due to summer heat and low snowfall levels concentrating the non-volatile components of the ice. Zeller et al. claimed that two major SEP events are also visible in the snow/ice record, indicated by two sharp peaks in the data. One of these events is the August, 1972 SEP event (again), the other event being from April, 1984. Certainly, the August, 1972 event did produce a major change in the atmospheric NOx content, and as we have already seen, this effect persisted for around 1 year. Zeller et al. suggest that the effects of these events could have been enhanced (in the snow record) by reductions in the snowfall at the time of the deposition of nitrate ions. The data for the 1972/3 southern summer shows a sharp peak in December, some 4 months after the event. This time delay represents the transportation time for the nitrates ions to reach the lower atmosphere and eventually to be removed as snow/rain. Dreschhoff & Zeller [14] later extended this record back to 1927 detecting evidence for further SEP events in July, 1946 and July, 1928. The event in 1946 was already a well-known particle event and the one in 1928 occurred around the time of a white light flare on the Sun. It is interesting to note that the events of 1972, 1946 and 1928 all occurred during the periods of total darkness at the south pole and represent increases of 7, 11, and 4 standard deviations above the series mean. When the 1928 event is corrected for the snow compactness, it increases to 6 standard deviations in significance. This process of using snow/ice records to study past SEP events can be extended even further back in time. The Greenland ice plateau is another place on Earth that preserves such a frozen record of past events. Dreschhoff & Zeller [15] and Kocharov, Ogurtsov & Dreschhoff [16] have analysed an ultra-high resolution ice core from Greenland. The data comes from a 122 metre long (10cm diameter) ice core drilled into the central East Greenland high ice plateau in summer 1992. The data clearly shows a series of large anomalies in nitrate ion concentration that are almost certainly due to solar particle events. They also measured the conductivity of the ice and this data shows volcanic activity that can be used to “date” the core. For events like Krakatau or Tambora, which are in the southern hemisphere, it is necessary to allow around 1 year for ion transportation to the northern polar regions. After removing the seasonal background from the nitrate data, it is possible to analyse the longer- term trends in the time series. By smoothing the data with a moving average with a length of about 8 years, several features of the series become apparent. There is a small dip in the early 1800s and a second longer dip from ∼1650Ð1700. These features correspond (in time at least), to the Dalton and Maunder Minima periods (respectively). These are two intervals when solar activity was at a reduced level. Indeed, during the Maunder Minimum, sunspots almost completely disappeared from the Sun’s surface and several solar cycles had only one or two spots in total. When directly compared to the sunspot data, the nitrate ion data does not correlate all that well. The lack of quantitative agreement between the two data sets reflects the fact that sunspots are not a good measure of solar flares and solar energetic particle events or the solar wind — which are the two dominant processes that drive nitrate generation (via SEPs and GCRs, respectively). However, when the data is examined at higher time resolution (i.e. without the smoothing), then some general agreement between the two sets can be found. Two peaks in particular occurred in 1851 and 1849. These could be connected to unusual white light flares seen in Feb., 1851 and Jan., 1849 — 1849 is close to sunspot maximum and 1851 is on the declining phase of the solar cycle. This is quite typical, the largest SEP events generally occur on the rising and declining parts of sunspot cycles - rather than at the peaks of the cycles. When the full data series is analysed for periodicities it is found that a ∼5year period is detectable in the data from 1760—1900. This period is almost exactly half the sunspot cycle and is very compelling evidence for an SEP signature to be present in the data (i.e. the data is likely detecting the rising and falling parts of each ∼11 year cycle). This conclusion was tested by taking the sunspot data and frequency doubling it (by multiplying the data by its Hilbert transform). This new sunspot series then showed a period at ∼5.3 years, exactly in agreement with the nitrate ion data. 6. CONCLUSIONS Galactic Cosmic Rays deposit their energy in the low stratosphere and produce NO at ratios depending on the phase of the solar cycle — this accounts for approximately half the NO at high latitudes and with alarge∼ 11-year modulation. Solar Energetic Particles Events occur sporadically but are more frequent around solar maximum. They interact with the atmosphere in the 30-60 km range but occasionally reach down below 20 km. However, they are useful for studying the effects of energetic particles on atmospheric chemistry because they have essentially instantaneous rise times with very large proton flux increases. Lightning is also a potentially important source of NO and is also possibly correlated with GCRs. Ice core data shows that SEP generated nitrates can reach the ground/low atmosphere in large quantities. So, some 40+ years after Ney’s original paper [1], what progress (if any?) have we made in in- vestigating the role that cosmic rays (or the resultant atmospheric ionisation) might be having on climate and/or meteorology? Certainly, the correlation reported by Svensmark & Friis-Christensen [17] between global cloud cover and galactic cosmic rays gave renewed vigour to the solar-activityÐclimate debate. If the link is correct it actually goes in the opposite sense to Ney’s original suggestion (figure 1). The exact details of the observational link between cloud cover and cosmic rays was refined when better cloud data became available (Marsh & Svensmark [18]). This showed a very striking correlation between low clouds and cosmic rays and in particular for low cloud top temperature which on the global correlation maps shows a clear and strong positive correlation for clouds in the Tropics. This latter links (i.e. low clouds and Tropics) then suggests a possible link with lightning which was shown above to also have an association with the Tropics and to be the only NOx process to operate below 5 km. Therefore, I would like to conclude this overview with my own updated “Ney Diagrammatic Scenario” (figure 6). Once again, like Ney [1], I would like to stress that this is only my suggested series of links and more Fig. 6: A revised Ney “diagrammatic scenario” showing a series of possible links between solar activity and cloud microphysics, based on some of the results and suggestions presented in this paper. The link between atmospheric conductivity and lightning needs some further testing and the final link between NOx species (and HNO3 in particular) and cloud processes is testable in the proposed CERN/CLOUD experimental facility. observations and experiments are needed to elucidate the various possible links in the chain of events shown. In particular, the actual demonstration of a total lightning frequency modulation with GCRs is lacking (primarily because of the lack of a good long term monitoring network/system for global light- ning statistics) and the link between NOx chemistry (perhaps through the action of nitric acid [HNO3]?) and cloud microphysics is also speculative. However, this last link is certainly easily testable in the proposed CERN/CLOUD experimental facility. ACKNOWLEDGEMENTS Iwould like to thank Jasper Kirkby, Henrik Svensmark and Mike Lockwood (and the other members of the Organising Committee) for inviting me to the First Ion-Aerosol-Cloud Interactions (IACI) workshop. Iwould also like to thank Robert Bingham for useful discussions and support for the work presented here. References [1] E.P. Ney, Nature, 183,(1959) 451. [2] P. Warneck, JGR, 77,(1972) 6589. [3] M. Nicolet, Plan.Spa.Sci., 23, (1975) 637. [4] C.H. Jackman, J.E. Frederick & R.S. Stolarski, JGR, 85, (1980) 7495. [5] M.R. Legrand, F. Stordal, I.S.A. Isaksen & B. Rognerud, Tellus, 41B, (1989) 413. [6] C. Jackman & S.R. Kawa, AGUSM, A62AÐ01 (2001) [invited review talk]. [7] C.T.R. Wilson, Proc. Roy. Soc. A, 92, (1916) 555. [8] M. Makino & T. Ogawa, JATP, 46 (5), (1984) 431. [9] B.A. Tinsley, J.Geomag.Geoelec., 48 (1), (1996) 165. [10] M. Lethbridge, GRL, 8,(1981) 521. [11] D.E. Olson, in Weather and Climate Responses to Solar Variations,ed. B.M. McCormac (Colorado Associated University Press, Boulder), (1983) 483. [12] R. Muhleisen,¬ in Electrical Processes & Atmospheres, eds. H. Dolezalek & R. Reiter (Steinkopff Verlag, Darmstadt), (1977) 467. [13] E.J. Zeller, G.A.M. Dreschhoff & C.M. Laird, GRL, 13 (12), (1986) 1264. [14] G.A.M. Dreschhoff & E.J. Zeller, Sol.Phys., 127, (1990) 333. [15] G.A.M. Dreschhoff & E.J. Zeller, Sol.Phys., 177, (1998) 365. [16] G.E. Kocharov, M.G. Ogurtsov & G.A.M. Dreschhoff, Sol.Phys, 188, (1999) 187. [17] H. Svensmark & E. Friis-Christensen, JATP, 59, (1997) 1225. [18] N. Marsh & H. Svensmark, Spa.Sci.Rev., 94, (2000) 215. EXPERIMENTS ON NUCLEATION PROCESSES IN AEROSOLS P. E. Wagner Institut für Experimentalphysik der Universität Wien, Vienna, Austria Abstract Various dynamical processes observed in the atmosphere are related to phase transitions from the gas phase. In this presentation selected experimental studies on nucleation and condensation processes are reviewed and their relevance to atmospheric aerosols is discussed. Open questions remain particularly for phase transitions in binary and multicomponent systems and in the field of ion-induced nucleation. Furthermore, the influence of miscibility and solubility of the compounds considered has not yet been clarified sufficiently. 1. INTRODUCTION Nucleation processes in the gas phase are of considerable importance in connection with formation and dynamical behaviour of atmospheric aerosols. While homogeneous nucleation in one-component vapor systems will generally not occur in the atmosphere, binary or multicomponent homogeneous nucleation are quite relevant to atmospheric aerosol formation [1-4]. Aerosol and cloud drop formation in the atmosphere are frequently connected with heterogeneous nucleation on soluble or insoluble particles and with ion-induced nucleation. The accuracy of nucleation experiments depends on the preparation of mixtures of inert gases with condensable vapors at well-defined thermodynamic conditions. This can be achieved by adiabatic expansion of initially saturated vapor-gas mixtures. In the present contribution expansion chamber studies on homogeneous and on heterogeneous nucleation will be discussed. The nucleated particles can be observed by the Constant-Angle Mie Scattering (CAMS) method [5]. The CAMS- detector allows non-invasive quantitative monitoring of number concentration and diameter of the condensing particles. 2. METHODS OF OBSERVATION A crucial condition for well-defined experimental studies of nucleation and condensation processes is the preparation of mixtures of an inert gas with condensable vapors at accurately determined supersaturations (vapor phase activities, partial vapor pressures) and temperatures. The following techniques have been successfully applied for this purpose: Nonisothermal vapor diffusion in static [6, 7] or steady state flow [8, 9] systems, adiabatic expansion [10, 11] and turbulent mixing [12] . As direct measurement of the temperature in supersaturated vapors is complicated, usually temperature and supersaturations need to be calculated. Nucleation processes can generally not be detected directly. Only the particles growing subsequent to the formation of the critical clusters, are observed. The concentrations of condensing droplets can be measured by single particle counting. The Constant-Angle Mie Scattering (CAMS) method [5] allows simultaneous and independent determination of concentration and size of growing droplets during condensation. 3. THE EXPANSION CHAMBER METHOD Supersaturated vapor - carrier gas mixtures can be prepared in expansion chambers (see, e.g., [13]). To this end a saturated or nearly saturated mixture of the gaseous components considered is adiabatically expanded in an expansion chamber. Corresponding to the adiabatic temperature drop occurring in the chamber, supersaturated vapors with well-defined vapor saturation ratios are obtained. An accurate straightforward determination of temperature drop and vapor saturation ratio is possible, if the carrier gas - vapor mixture is nearly an ideal gas and if no significant vapor condensation occurs already during the expansion (dry-adiabatic expansion). Approximately dry- adiabatic expansions can only be achieved, if the time interval, during which the expansion occurs, is small compared to the typical times required for condensational growth at the conditions considered. For expansion chamber studies of homogeneous nucleation a further condition is essential. As homogeneous nucleation and droplet growth generally occur simultaneously, the depletion of vapor caused by the condensational drop growth process will lead to a reduction of vapor supersaturation thereby quenching the homogeneous nucleation process. In order to obtain quantitative information on homogeneous nucleation rates, it is therefore important to decouple nucleation and growth. This can be achieved by means of the nucleation pulse technique [10, 11]. The droplets growing in the expansion chamber can be quantitatively observed by means of the CAMS detection method [5]. To this end the droplets are illuminated by a laser beam. The light flux transmitted through the expansion chamber as well as the light flux scattered at a selectable constant scattering angle are monitored simultaneously. The scattered light flux shows series of extrema in good agreement with the prediction by Mie theory (see [5]). After establishing a unique correspondence between experimental and theoretical light scattering extrema, size and number concentration of the growing droplets can be determined by quantitative comparison of experimental and theoretical light fluxes. Light attenuation in the expansion chamber can be accounted for by normalizing the scattered relative to the transmitted light fluxes. The following features of the expansion chamber method are notable: (1) Temperature and vapor saturation ratios (vapor phase activities) in expansion chambers can be determined accurately by means of a straightforward procedure, (2) Vapor supersaturation can be achieved during a comparatively short time interval, (3) Temperature and vapor saturation ratios obtained in the measuring volume are uniform, (4) Number concentration and size of growing droplets can be measured at various times during the growth process by means of a non-invasive and absolute method (CAMS detection). 4. EXPERIMENTAL RESULTS Homogeneous nucleation in unary vapors has been studied for various compounds generally showing fair agreement with theory. Particularly the experimental slopes of the nucleation rate vs. supersaturation curves agree well with theory. However, experiments for binary and ternary vapor mixtures [14-19] frequently result in substantial deviations from classical nucleation theory, particularly for non-ideal mixtures. Homogeneous nucleation of partially immiscible liquids shows a somewhat complex behaviour. The few so far available quantitative experiments on heterogeneous nucleation have been restricted to unary systems. Only recently we have reported first experimental studies on binary heterogeneous nucleation [20]. These studies show that the macroscopic contact angle is hardly applicable for heterogeneous nucleation on nano-particles. 5. CONCLUSIONS Expansion chamber experiments are suitable for time-resolved measurements of nucleation and condensation processes. For the purpose of modeling of atmospheric conditions expansion chambers have the important feature that the thermodynamical conditions in the measuring volume are uniform and can be accurately determined. Experiments on homogeneous nucleation typically show fair agreement with the classical nucleation theory for unary systems, whereas severe discrepancies for some multicomponent systems have been encountered. The interpretation of heterogeneous nucleation measurements depends on the choice of the contact angle. ACKNOWLEDGMENTS This work has been supported by the Fonds zur Förderung der Wissenschaftlichen Forschung, Proj. Nr. P 9421 and by the Hochschuljubiläumsstiftung der Stadt Wien. REFERENCES [1] Raes, F., and Van Dingenen, R., J. Geophys. Res. 97, 12901 (1992). [2] Arstila, H., Korhonen, P., Kulmala, M., J. Aerosol Sci. 30, 131 (1999). [3] Mäkelä, J. M., Aalto, P., Jokinen, V., Pohja, T., Nissinen, A., Palmroth, S., Markkanen, T., Seitsonen, K., Lihavainen, H., and Kulmala, M., Geophys. Res. Lett. 24, 1219 (1997). [4] Kulmala, M., Pirjola, L., and Mäkelä, J. M., Nature 404, 67 (2000). [5] Wagner, P. E., J. Colloid Interface Sci. 105, 456 (1985). [6] Katz, J. L., and Ostermier, M., J. Chem. Phys. 47, 478 (1967). [7] Hung, C., Krasnopoler, M., J., and Katz, J. L., J. Chem. Phys. 90, 1856 (1989). [8] Anisimov, M. P., and Cherevko, A. G., J. Aerosol Sci. 16, 97 (1985). [9] Wilck, M., Hämeri, K., Stratmann, F., and Kulmala, M., J. Aerosol Sci. 29, 899 (1998). [10] Schmitt, J. L., Adams, G. W., and Zalabsky, R. A., J. Chem. Phys. 77, 2089 (1982). [11] Wagner, P. E., and Strey, R., J. Chem. Phys. 80, 5266 (1984). [12] Wyslouzil, B. E., Seinfeld, J. H., and Flagan, R., C., J. Chem. Phys. 94, 6842 (1991). [13] Wagner. P. E., in Aerosol Microphysics II (W. H. Marlow, Ed.) p.129. Springer, Berlin, 1982. [14] Wilemski, G., J. Phys. Chem. 91, 2492 (1987). [15] Schmitt, J. L., Whitten, J., Adams, G. W., and Zalabsky, R. A., J. Chem. Phys. 92, 3693 (1990). [16] Wyslouzil, B. E., Seinfeld, J. H., Flagan, R. C., and Okuyama, K., J. Chem. Phys. 94, 6842 (1991). [17] Strey, R., Viisanen, Y., and Wagner, P. E., J. Chem. Phys. 103, 4333 (1995). [18] Viisanen, Y., Wagner, P. E., and Strey, R., J. Chem. Phys. 108, 4257 (1998). [19] Viisanen, Y., and Strey, R., J. Chem. Phys. 105, 8293 (1996). [20] Petersen, D., Ortner, R., Vrtala, A., Laaksonen, A., Kulmala, M., and Wagner, P. E., J. Aerosol Sci. 30, S35 (1999). COMMENTS ON THE OPERATION OF A WILSON EXPANSION CLOUD CHAMBER John L. Schmitt Physics Department and Cloud and Aerosol Sciences Laboratory, University of Missouri-Rolla, Rolla, MO, 65401 USA Abstract Comments are made concerning the practical operation of a Wilson expansion cloud chamber for experimental research in heterogeneous and homogenous nucleation. Topics covered are the general operation of the chamber, the detection of cloud droplets, the cleaning of the chamber and the interpretation of nucleation data. The comments include successful and unsuccessful experimental techniques. 1. INTRODUCTION The main experimental apparatus proposed for the CLOUDS investigation is a Wilson expansion cloud chamber. The following comments are intended to provide guidance in the operation of that chamber by examining what has been done while operating an expansion chamber at the Cloud and Aerosol Sciences Laboratory at the University of Missouri-Rolla. The comments are based on about 25 years experience by the author operating a chamber (designed for homogeneous nucleation experiments). This article is intended to supplement Refs. [1, 2] by the author and other publications, e. g. Refs. [3, 4]. Reference [1] is a description of the cloud chamber as constructed and used by the author. While the basic features of that chamber have not changed, this article does address some of the evolutionary changes that have taken place since it was constructed. Reference [2] describes a Wilson cloud chamber, the UMR absolute Aitken nucleus counter, that was specifically built to detect particles that nucleate from near 100% Relative Humidity to beyond the ion limit. That chamber no longer exists. Reference [3] is a review paper that addresses the “Wilson Cloud Chamber and Its Applications in Physics” circa 1946 and still contains much valuable information. Reference [4] is a book by J. G. Wilson (not C. T. R. Wilson) which reviews the Wilson cloud chamber circa 1951. The following comments include improvements, attempted improvements, suggestions, speculations and failures. Again this article primarily addresses experiments with the author’s apparatus, and therefore the references and anecdotal evidence are greatly skewed towards work done at high supersaturation with that device. 2. GENERAL OPERATION 2.1 Chamber principles The physical operating principle of the expansion chamber is that if a gas, containing a condensable vapor at equilibrium, is “suddenly” expanded, the temperature of the gas will adiabatically decrease. In the following, the background carrier gas in the chamber will be referred to as the gas and the condensable vapor as the vapor. At the lower, expanded, temperature the equilibrium vapor pressure of the vapor is much lower and since the vapor has not had time to diffuse, it is present at a higher concentration than equilibrium, i. e. supersaturated. Thus the expansion chamber is a device that produces a supersaturation upon demand (expansion). Formally, supersaturation ratio is the ratio of the actual vapor pressure in the chamber divided by its equilibrium vapor pressure at the given temperature but will be referred to in the following as only supersaturation. The length of time for which one knows, to the required accuracy, the physical conditions, temperature and vapor content, in the central volume of the chamber during and after the expansion, is controlled primarily by the transit (time) of heat and vapor from the walls. The material or substance in the liquid pool in the case of homogenous nucleation is determined by the experiment. For atmospheric simulation water would be the material of choice. For ion nucleation, the classical mixture, used to detect ion tracks for nuclear physics, is about 1 part water and 2 parts ethyl alcohol, Ref. [3] pg. 237, which produces the smallest expansion ratio for the mixture to nucleate on ions. 2.2 Physical parameters in the chamber Initial conditions in the chamber are temperature, pressure and vapor content of the gas. Temperature and pressure are measured by sensors and the vapor content of the carrier gas is known from the equilibrium vapor pressure of the liquid pool or by flushing with gas of known temperature and vapor content Ref. [5]. The conditions during and after the expansion are calculated from initial temperature and pressure, the expanded pressure and the thermodynamic properties of the gas and the vapor. Pressure equilibrates at the speed of sound, however note that the speed of sound is not constant, thus pressure can be a valid parameter to describe the conditions in the experimental volume of the chamber. Direct measurement of the temperature in the experimental volume is very difficult since a sensor larger than molecular size in the central volume will probably induce heterogeneous nucleation, along with latent heat release. Laser spectroscopy of a carefully selected molecule has been suggested as method of temperature measurement Ref. [6] but this suggestion has not been implemented. Problems lie in the areas of isolating and measuring molecules only in the central experimental volume and finding a molecule that not only has the desired temperature range and sufficient sensitivity to measure temperature to 0.1 C but also will not interact with the desired nucleation. However, such a molecule could be present in large numbers and this application of laser spectroscopy does not have the same problem as attempting to apply laser spectroscopy to nucleation embryos that are present in very low concentrations. A typical temperature calculation ranges from simple, for an adiabatic expansion in a perfect gas with a minor nearly ideal vapor component (the calculation primarily is the integration of specific heat as a function of temperature), to complex, for a real gas with a significant pressure contribution from real vapor. Argon works well for the gas: it is ideal and it does not leak through seals easily. Helium leaks easily, has a high heat conductivity and new gas cylinders of helium often produces alpha particle tracks Ref. [7]. Air is not an ideal gas but would be needed for simulating the atmosphere. Additional complexity is added if heat and vapor from the chamber walls reach the experimental volume during the experiment and droplet growth, with attendant latent heat release and vapor depletion, occurs. In each case accurate thermodynamic data for the gas and vapor is essential. If one considers the best case of an ideal gas and a low vapor pressure vapor, the gas mixture is very close to ideal since the contribution to the specific heat by the vapor is small. This also means that the thermodynamic data for the vapor can be uncertain since its contribution is small. In addition, the mixture will produce a maximum temperature change on expansion. 2.3 Nucleation pulse technique The nucleation “pulse” technique is often used: the chamber is expanded quickly to maximum expansion, is then recompressed slightly and held at a supersaturation lower than the supersaturation at the maximum expansion. Since (homogeneous) nucleation is an exponential process as a function of supersaturation, a slight reduction of supersaturation lowers the nucleation rate by orders of magnitude and thus effectively, when the nucleation rate is integrated over time to calculate the total number of droplets, “all” nucleation occurs very near the maximum supersaturation. (In practice one determines the nucleation rate, droplets/cm3 sec, from the droplet count, droplets/cm3, and the integrated pulse length.) This also limits the nucleation to a known time and length of time. The droplets grow to detectable size only after the slight recompression. This limits the latent heat release and vapor depletion due to growing droplets to the time after nucleation and thus conditions in the experimental volume are well known during nucleation. The nucleation pulse technique avoids the problems of simultaneous nucleation and growth. However, these effects will have to be included in the models if the experiment requires that nucleation and growth take place simultaneously. 2.4 Supersaturation range and detection of low concentrations The expansion chamber can produce a wide range of supersaturation. Water supersaturation in the range from 1.00 (100% Relative Humidity) to 1.05, nucleates on and detects (aerosol) particles, i. e. heterogeneous nucleation. Much higher supersaturation, e.g. 3.5 and up, will nucleate on and detect negative ions and even higher supersaturation will homogeneously nucleate water (examples are for expansions from about 300K, initial chamber temperature). The chamber is capable of detecting by droplet growth, particles from microns in size (particles at the lower limit for easy detection by Mie light scattering and which do not fall out rapidly) to molecules. For example, if one has an molecule, ion, or particle that will nucleate at a lower supersaturation than the homogeneous nucleation of the background vapor, then the chamber is able to nucleate and grow a droplet on that molecule, ion or particle. For reference, a gas at one atmosphere pressure has about 1019 atoms or molecules/cm3, and one droplet/cm3 can easily be detected by nucleation in a cloud chamber Ref. [8]. 2.5 Low supersaturation expansions The expansion chamber was chosen as a detector for CLOUDS because it can produce a wide range of supersaturation. It has been traditionally used at high supersaturation in homogeneous nucleation experiments. One technique to produce low supersaturation is to introduce undersaturated vapor and then expand that mixture, Ref. [5]. If, on the other hand, the equilibrium vapor pressure of the liquid pool is used to provide the vapor, a mixture in the pool can produce (an application of Raoults’ Law; one also must know activity coefficients) undersaturation in the vapor. In the case of water, a mixture of tetraethylene glycol and water will initially undersaturate the chamber in water and require large (volume ratio) expansions to produce saturation or supersaturation.Ref. [9]. 2.6 CLOUD chamber The considerable precision with which it is anticipated that the piston of the proposed CERN cloud chamber can be controlled, introduces a capability not found in previous expansion cloud chambers. One now can consider not only experiments where the chamber is expanded once to a high or low supersaturation to nucleate on particles or ions, but also experiments where the position of the piston is accurately controlled with time to attempt to program the Supersaturation (pressure) with time. For example, the chamber is expanded to produce a supersaturation high enough to nucleate on an ion and grow a droplet, the chamber is then compressed to evaporate the droplet and a second expansion is used to detect any re-evaporation nuclei that remain. Experiments of this type have been performed using the UMR simulation chamber Ref. [10]. That chamber has the additional and significant advantage of exact control of the wall temperature with time (up to 10 C/minute rate of change with 0.02 C accuracy). One can match wall temperature to the gas temperature during the experiments. Obviously experiments of this type in the CLOUDS chamber will require good models for the transport of heat and vapor and droplet growth. Finally the motion of the piston in the proposed CLOUDS chamber does not depend upon a difference in pressure across it, to drive an expansion. The hydraulic cylinder and valve system has considerable force and pulls the piston. Therefore, the CLOUDS chamber can be used with very low pressures in the experimental volume (for the simulation of high altitudes). However, one needs to be very careful to model these experiments. For example, accurate calculations for an adiabatic expansion in a low pressure gas that is mostly water vapor at very low temperatures (the water vapor may be supercooled) and pressures require accurate thermodynamic data. This may be difficult to find or measure. 3. DETECTION OF DROPLETS 3.1 Droplets vs. nucleation embryos One detects droplets in the chamber and equates the number of droplets to the number of embryos that were nucleated. This assumption should be valid if the number of embryos is not large enough to coagulate before or during growth or embryos are not destroyed or created (after the nucleation event). All (current) methods of detection do not detect the embryo, perhaps 10 to 50 molecules, directly but depend upon growing it to a size large enough to detect. At present, very small particles have been examined by neutron scattering Ref. [11]. In principle it might be possible to “interrogate” by laser spectroscopy the nucleation embryos or ions, however one must also note that typically one has 100 embryos/cm3 in a gas at atmospheric pressure with approximately 1019 atoms or molecules/cm3. Finally, ion mobility experiments have been anticipated in the CLOUDS chamber by including a field cage in the design. 3.2 Light scattering detection The most direct detection methods allow the droplets to grow to about 2 microns (radius) and larger and use visible light to “see” the droplets, e.g. photography and Mie scattering. A straightforward technique is to select droplets using a “sheet” of light 1 cm thick from xenon flash tubes and photograph, from a viewpoint at right angles to the sheet of light, with a high resolution lens and image detector. One calibrates the system by photographing a 1 cm standard grid. A 1 cm2 projected area on the detector images a depth in the light sheet of 1 cm and thus one counts droplets in 1 cm3. Most uncertainty in the measurements comes from the “thickness” of the edge of the sheet of light. The basic concept is that the illumination defines the volume that is measured. Images on the detector usually do not measure the size of the droplets but are only a result of the resolution limitations of the film and lens combination. Photographic film has the advantage of very high information content and storage and but also requires handling and processing. The advantages of an electronic detector, e. g. a CCD, are that the image is available very quickly, which is essential in experiments that require interaction, and the image is digital, which can be immediately analyzed by a computer. The disadvantages are that CCD sensors are expensive in the large sizes, e. g. 4000 by 4000 pixels (a 35-mm camera frame can be considered to have approximately 2400 by 3500 pixels for a frame size 24 by 35 mm with 10 micron resolution) and it is difficult to understand what resolution really means for the square or rectangular pixels in a CCD. For example, what is the resolution along the diagonals of the pixels or what can be said about a droplet that images at the intersection of four pixels? An example of a photographic system is a 1500 by 1000 pixel CCD with 9-micron square pixels combined with a camera lens that will resolve 125 line pairs/mm using the sheet of light technique to illuminate the cloud droplets. The lens is used at f/4, producing a depth of field, i.e. all droplets in acceptable focus, of at least 1 cm (the thickness of the flash tube sheet of light). The CCD detector covers about 25% of the 35-mm film image that it replaced. 3.3 Image analysis Analysis of the photograph or image can be done by computer image analysis or by visual counting. Visual image analysis (one looks at the displayed image on a computer screen with a superimposed 1 cm2, scaled, grid and counts the droplets by eye) has the advantage that the eye and brain are very good at distinguishing artifacts and multiple drops. For very low counts, 1 drop/cm3, visual counting can distinguish flaws that if included would greatly skew the drop count. For example, expansions to supersaturations above the ion limit will probably detect cosmic rays. Cosmic rays do not produce the traditional tracks but usually produce “blobs” which are the result of a cosmic ray coming through the chamber from above (perpendicular). One thus sees a section of the perpendicular track and the blob is produced when the ions in the track diffuse and are detected as ions. For very high counts, 1000 droplets/cm3, the eye is good at distinguishing multiple droplets and droplets close together. For experiments visual image analysis is a tedious but straightforward exercise. The error in the drop count is considered to be the square root of the number of drops counted. For concentrations of 1 droplet/cm3, one typically must count as many cm3 as possible on the image. Attempts to count images with computer programs from several sources have produced very mixed results. Some of the results of these attempts: The image of the “droplet” should cover approximately 6 or more pixels so that one can use a measure of the roundness of the image and the number of pixels/droplet to validate a droplet and to distinguish between multiple droplets Ref. [12]. In practice having a droplet cover 6 pixels also limits the CCD camera field of view in the chamber. The photographic system described above has about 1 to 2 pixels/droplet. In general, automatic drop counting systems appear to work well only restricted range of droplets concentrations. Often computer assisted drop counting took longer to do than counting by eye and required careful and constant evaluation for accuracy. 3.4 Mie scattering detection Mie visible light scattering from droplets is another means of detection. At visible wavelengths one begins to detect droplets at about 2 microns in size. Using white light as illumination allows one to use the relationship between scattered intensity and size that is roughly a straight line from 2 to 15 microns. A (Mie) scattering system, for which the calculations were done but which was not constructed, which has the potential to measure to measure size and droplet concentration by scattered intensity, is to use all four strong laser lines (combined) from an argon ion laser as an approximation to white light. This “white” light laser system could be used, to analyze clouds that are polydispersed. This system for counting droplets in real time in the chamber uses a three- mirror scanning system, X, Y and Z, that raster scans a focused (white light) laser beam over 1 cm3 in the center of the chamber. The focused laser beam isolates the droplets in the desired location. The droplet is detected by the pulse of light scattered as the droplet is scanned by the beam. The intensity of the pulse is proportional to the droplet’s size. The major problem with this system is that while it was concluded that it would work well with a few hundred droplets/cm3 (each relatively isolated), it would not work well at only 1 droplet/cm3 (a much larger volume would need to be scanned to obtain good accuracy), and multiple scattering might be a problem in the 1000 droplets/cm3 range. It however would detect and size each droplet. That information could be used to identify the large droplets that have most of the liquid water content of a cloud. The detection of size as a function of time can be done with Mie scattering using monochromatic radiation scattered from monodispersed droplets. The CAMS, Constant Angle Mie Scattering, system for the proposed CLOUDS chamber is described in the CLOUDS proposal, Ref. [13]. In general, one uses a collimated polarized laser beam in the chamber and detects the scattered light at a fixed angle. As the droplets in the (monodispersed) cloud grow, the Mie scattering peaks (lobes) move past the detector. The resulting intensity pattern as a function of time is compared to theoretical calculations of size vs. intensity to find size vs. time. The accuracy of the technique is about 0.1 micron. Careful selection of the angle of scattering can greatly simplify data analysis. If one selects, e. g. 30 deg, the intensity vs. time is a pattern that is similar to a sine wave with a D. C. component increasing with time. If the detection of the first peak in the signal is questionable due to noise, it is very difficult to determine which peak corresponds to which size on the theoretical curve. However, if one uses angles near 4 degrees, the scattering pattern is a modulated signal and the modulation allows one to identify Mie peaks easily, even if some of the low intensity peaks are lost in noise. In addition, the scattering pattern changes very rapidly with angle in the 4 to 5 degree range, which allows one to measure a scattering pattern and determine from it the scattering angle experimentally, a great convenience. This system will also measure the number of droplets if scattered light intensity can be directly related to the scattered light intensity from a single droplet. Note that these detection systems are for clouds with droplets of one size, i. e., monodispersed. A cloud with droplets of many sizes is much more difficult to characterize. For example, the intensity of Mie scattering from a droplet using a monochromatic laser beam varies greatly with very small changes in size and, e.g. a 5.1-micron droplet may scatter much more light than 9.6-micron droplet. This latter property also makes it very difficult to deconvolute the total scattering signal from a cloud and determine the droplet size distribution. 3.5 Detection of ice The detection of ice in the chamber requires a different technique than the detection of droplets. Light (Mie) scattering calculations are difficult since ice crystals are not spheres. Figure 19 in the CLOUDS proposal, Ref. [13], illustrates photographic detection in our cloud chamber. That illustration shows a cloud of supercooled water droplets, at about - 40 C, in which some droplets have frozen. One observes that the ice particles scatter much more light, appear brighter in the photographs, and thus are easily identified. It is assumed that the ice has a polycrystalline structure because the ice images are all the same intensity, as opposed to, e. g. a needle or a plate, which would reflect light much differently depending upon its orientation with respect to the illumination. No further investigation was made on this assumption and this work is the only work that has been done with our chamber on ice. It is appropriate to mention that the Karlsruhe aerosol chamber discussed at the IACI meeting, Ref. [14], has an ice detection system based on depolarization in the observed backscattering of a polarized laser beam from the ice crystals. 4. TEMPERATURE CONTROL Accurate control and knowledge of the temperature of the chamber is vital. See Ref. [1] and the CLOUDS proposal, Ref. [13], for chamber temperature control details. In addition, the chamber has been operated with a gradient of typically 0.3 C from top to liquid pool to inhibit convection and the windows are controlled at temperatures slightly higher than the gas and vapor to prevent condensation on them 5. CHAMBER CLEANING 5.1 Solvents The chamber has been primarily used for homogeneous nucleation experiments and therefore it is essential that it be cleaned as well as possible. Most probably the cleaning technique used on the CLOUDS chamber will depend greatly on the experiment being done. For example, an experiment that routinely admits outside gas will have different requirements than an experiment using ion detection in a sealed system. In addition to the cleaning and the attention paid to the removal of trace contamination of the gas, described in Ref. [1], the chamber walls are now cleaned before experiments by flushing solvents over the walls. Typically the material from the previous experiment is removed by draining, by suction and by evaporation. A fog nozzle is placed in the center of the chamber and, about 10 liters of solvent are forced, using gas from the same system that supplies gas for the experiments, though the fog nozzle to produce a cloud of droplets. The droplets impact on the chamber walls, the resulting liquid runs down the walls and collects in the bottom. Solvents used include water, ethyl alcohol, acetone and some of the material for the next experiments. Water works well in certain experiments, is readily available and easily disposed of. Acetone is very effective for removing hydrocarbons. One also must be careful that the solvents do not introduce contamination. For example, carbon tetrachloride was found to contain a stabilizer that leaves a film behind when it is evaporated. One must also be very careful in using evaporation to remove material. Some materials, such as water, easily develop a surface film that inhibits evaporation very effectively. In addition, homogenous nucleation can be extremely sensitive to trace molecules and we have found that “high purity” materials from different chemical suppliers vary widely with respect to nucleation properties in experiments. Materials also will oxidize and thus may change nucleation properties with time in the chamber, if a gas other than argon is used, or improperly stored. One technique for filling the liquid pool in the chamber is to evacuate the chamber and pull the material in from the original bottles. One can flush the bottle with clean nitrogen. Finally, when the operating liquid is put in the chamber, it is wise to use a vacuum and pressure cycle to remove dissolved gas and high vapor pressure impurities from it. If necessary the gas in the chamber can be cleaned by dilution. One can use repeated cycles of using the piston to compress the gas followed by removal by vacuum. 5.2 Trace impurities and re-evaporation nuclei The chamber is constructed of stainless steel and glass with fluorocarbon seals and was extensively cleaned after its construction to rid it of such pollutants as cutting oil from machining. After it was constructed, it was tested with water nucleation. The new chamber produced homogeneous water nucleation data that was the same as the previous chamber Ref. [15]. This was considered necessary and sufficient evidence that the new chamber was working correctly and furthermore that the experimental technique cleaned, see below, the chamber and removed trace impurities from the liquid, vapor and gas. However, several years later the following experiments were done. The chamber was cycled to produce homogeneous nucleation of water and the droplet growth rates were measured used Mie scattering. The chamber was then cleaned again, this time using about 10 liters of reagent grade acetone that was fogged onto the walls. Finally, the chamber was cycled for water nucleation and the droplet growth rates were measured. They were about a factor of 3 faster. The water nucleation data was the same as before. This was considered evidence that a surface layer, due to trace contamination (hydrocarbons were suspected) was on the droplets in the previous experiments. Subsequent experiments with the UMR simulation chamber showed that a condensation-evaporation-condensation cycle produces considerable changes in the condensation coefficients Ref. [16]. A contaminated surface on the droplet was postulated. It is important to note that the sticking coefficient for a water molecule on the surface of a droplet may greatly differ from the case of a clean water surface to the case of the “dirty” surface the droplet may have in the real atmosphere. This is particularly true if the droplet has been processed through a condensation- evaporation-condensation cycle, e.g. in a cloud. 5.3 Chamber self-cleaning The above description of how to clean the chamber primarily utilizes solvents to dilute and remove impurities. This may be sufficient for experiments, e.g. with aerosols, at low supersaturation. Only operation of the CLOUDS chamber will answer that question. However, for high supersaturation to detect ions and possible formation of molecular sized nucleation embryos, it is essential to use the self-cleaning properties of the chamber. In fact one may state, with justification, that the reason the Wilson expansion cloud chamber, with its ability to detect 1 ion/cm3 in a gas at atmospheric pressure, can be used at all is that it is self-cleaning. For a typical data cycle, the basic idea is to expand the chamber to deep expansions that nucleate on impurities, grow droplets, and rain out the droplets. The assumption is that the expansion rains out the impurities and traps them in the liquid pool. A typical experiment, at the time of publication of Ref. [1], was to expand the chamber to a supersaturation slightly lower than the supersaturation needed for the experiment, e.g. homogeneous nucleation, and nucleate on the impurities. The “impurities” includes re-evaporation nuclei, i. e. something left behind when the droplets from the previous data expansion evaporated, Ref. [4], pgs. 2 and 13. A typical experiment included, the first cleaning expansion in which one observed approximately 10 droplets/cm3, a 5-minute wait to equilibrate, the second cleaning with 1 droplet/cm3, a 5-minute wait, the third expansion with very few droplets visible in the entire chamber, a 5-minute wait and finally the deep data expansion to a higher supersaturation for homogeneous, nucleation. This technique was used to measure the homogeneous and ion nucleation for water and several other substances. It produced data that was repeatable and internally consistent. 5.4 Homogeneous nucleation data as a lower bound Experiments on the nucleation of octane isomers produced results that raised questions about the basic assumptions made about cleaning the chamber by repeated expansions Ref. [17]. It was found that if one used cleaning expansions to a slightly higher supersaturation than the supersaturation reached in the data cycle, that the number of droplets was less, in the subsequent data expansion, than if the cleaning cycles were to slightly lower supersaturation than the data expansion (previous technique). In addition, re-evaporation nuclei were not observed. It was also observed that running the chamber over a long period of time, weeks, in several cases produced dramatic changes in the supersaturation (higher) measured for “homogeneous” nucleation. While this is not yet completely understood, we propose the following model. The high supersaturation expansions nucleate on the trace impurities (as well as homogeneously nucleate a few droplets) and rain out the droplets into the liquid pool. Since many more droplets are formed in the deep expansion, large amounts of impurities are subsequently rained out. However while some of these impurities stay in the liquid pool others have significant vapor pressures and diffuse back into the gas. Since the chamber must come to thermal equilibrium before the next expansion, there is time for re-diffusion of the impurities. In the case of octane we observe no re- evaporation nuclei and thus we can expand to deep expansions without creating additional “impurities” with the nucleation process. Materials, at this time all appear to be hydrogen bonded such as water, that produce re-evaporation nuclei must be investigated with the original technique. Recent experiments on water, Ref. [18], have produced data that shows fewer droplets at higher supersaturations than the first extensive date set on water Ref. [15]. These experiments were performed with full knowledge of improved cleaning techniques (plus more experience and hindsight). Our general conclusion is that nucleation data is be represented as a lower bound. The appropriate conclusion to draw from this for the CLOUDS chamber is that, again, a Wilson cloud chamber is an extremely sensitive detector and that trace impurities can and will influence results significantly. The recent water data illustrates, unfortunately, that good data and bad data can look very similar. Particularly when there is no other data. 6. SUMMARY These comments supplement previous publications, Refs. [1] to [4], on the Wilson expansion cloud chamber. Again they are intended as practical suggestions on how to operate an expansion chamber for nucleation work and incorporate techniques that work suggestions and failures. ACKNOWLEGEMENTS The author would like to mention that the Cloud and Aerosol Sciences Laboratory at the University of Missouri-Rolla was founded by J. L. Kassner, Jr., who adapted expansion cloud chamber construction and technique from the area of nuclear physics to the study of nucleation. REFERENCES [1] J. L. Schmitt, Rev. Sci. Instrum. 52 (1981) 1749. [2] J. L. Schmitt, J. Aerosol Sci. 13 (1982) 373. [3] N. N. Das Gupta and S. K. Ghosh, Rev. Mod. Phys. 18 (1946) 225. [4] J. G. Wilson, The Principles of Cloud-Chamber Technique, (Cambridge University Press, 1951). [5] R. Strey et al., J. Phys. Chem. 98 (1994) 7748. [6] A. Schawlow, private communication. [7] J. L. Kassner, Jr., private commuication. [8] H. N. Ereifej et al, J. Appl. Phys. B68 (1999) 141. [9] J. Schmitt and D. Hagen, Atmospheric Res. 31 (1994) 91. [10] D. White et al., Rev. Sci. Instrum. 58 (1987) 826. [11] B. E. Wyslouzil et al., J. Chem. Phys. 113 (2000) 7317. [12] W. Walton, graduate student, private communication. 13] B. Fastrup et al., A study of the link between cosmic rays and clouds with a cloud chamber at the CERN PS (CLOUDS Proposal) (CERN, 2000). [14] O. Mohler, The Karlsruhe Aerosol Chamber Facility AIDA, Workshop on Ion-Aerosol- Cloud Interaction (IACI), CERN, 18-20 April 2001. [15] R. Miller et al., J. Chem. Phys. 78 (1983) 3204. [16] D. Hagen et al., J. Atmospheric Sci. 46 (1989) 803. [17] G. J. Doster et al., J. Chem. Phys. 113 (2000) 7197. [18] J. L. Schmitt et al., Nucleation and Atmospheric Aerosols, 15th International Conf., B. Hale and M. Kulmala, eds., 2000 (AIP Conf. Proc. 534) 51. THE KARLSRUHE AEROSOL CHAMBER FACILITY AIDA: TECHNICAL DESCRIPTION AND FIRST RESULTS OF HOMOGENEOUS AND HETEROGENEOUS ICE NUCLEATION EXPERIMENTS O. Möhler, A. Nink*), H. Saathoff, S. Schaefers, M. Schnaiter, W Schöck, and U. Schurath Institute of Meteorolgy and Climate Research, Forschungszentrum Karlsruhe, Germany Abstract The large experimental facility AIDA of the institute of meteorology and climate research at Forschungszentrum Karlsruhe is operated and used as a cloud chamber to study processes of ice formation in tropospheric and stratospheric clouds. Like in clouds, particle freezing and growth is initiated by expansion which leads to quasi-adiabatic cooling and thus ice- and water supersaturation at constant wall temperature. Intensity and depolarisation of forward- and back-scattered laser radiation is measured, caused by particles in a small scattering volume far from the walls. The ice phase is also detected by in situ FTIR spectroscopy. Number size distribution of interstitial aerosol and activated ice particles is measured with an optical particle counter. Various insoluble aerosol components can be generated and added to the chamber in order to investigate their influence on ice formation processes at controlled temperatures, cooling rates, and supersaturations. 1. INTRODUCTION Ice particle formation is an important process in the troposphere and stratosphere. It can occur either by homogeneous freezing of droplets below about Ð35∞C [1], or be heterogeneously induced by so-called ice nuclei. E.g. it is speculated that soot particles from aircraft can act as ice nuclei [2]. A quantitative description of these processes is crucial for a better understanding of the lifetime of clouds with respect to rainout, and their optical properties. Distinction between supercooled liquid and frozen aerosol particles (cloud hydrometeors; PSC particles) is essential for the investigation of these ice nucleation processes. Polar stratospheric clouds play a crucial role in the ozone destruction process. During recent years, various physical and chemical particle formation processes have been investigated intensively. Liquid ternary solution particles, crystalline hydrates, and PSCs mainly composed of ice or mixtures of liquid and solid particles have been detected and analysed by remote sensing [3, 4] and in situ techniques [5]. The ability of PSCs to induce chemical ozone depletion is a function of particle concentration, size, composition, and thermodynamic phase. For example, solid particles can grow bigger than liquid and thereby are thought to be responsible for denitrification of the lower stratosphere, enhancing and extending ozone depletion [6]. So far the formation processes of solid particles are not fully understood. 2. AIDA FACILITY 2.1 Experimental setup A schematic cross section of the AIDA cloud chamber and some analytical and technical instrumentation is shown in Figure 1. This chamber (Volume = 84 m3) is operated over a wide *) Now at Bayer AG, Leverkusen, Germany. range of atmospheric conditions: -90∞C < T < +60∞C, r.h. under static conditions near 100 %; pressures from above 1 bar to below 1 mbar; ice and water supersaturations. This covers conditions throughout the troposphere and lower stratosphere under which water clouds, mixed clouds, cirrus clouds, and even Polar Stratospheric Clouds (PSC) are formed. Temperature Controlled Housing oo -90 C to +60 C Aerosol Generator Aerosol Vessel AIDA Dew Point Filter Hygrometer Samples D M FTIR and UV/VIS Condensation A Spectrometer Nuclei Counter T, p Condensation Scattering and Nuclei Counter Depolarisation Detector 2 Detector 1 Argon Ion Laser 488 nm Scatering Depolarisation Expansion Optical Volume Particle Spectrometer Vacuum Synthetic Air Vacuum Cryostat Liquid Pump Supply Pump Nitrogen Fig. 1 AIDA main components and instrumentation available for ice activation experiments. 2.2 Ice activation experiments Experiments investigating ice formation at supersaturated conditions typically are started at homogeneous temperature conditions, pressure between 180 hPa and 1000 hPa, and relative humidity close to ice saturation controlled by an ice covered chamber wall. Ice supersaturation is achieved by volume expansion due to controlled pumping using two large vacuum pumps at different pumping speeds. Depending on starting temperature and pumping speed, the regime between ice and water saturation is passed within a few minutes at cooling rates up to 200 K/h. Figure 2 shows an expansion started at 174 hPa and 225 K. The expansion period lasted about 8.5 min. The highest cooling rates are only achieved within the first few minutes. Hereafter, a steady state is achieved between further adiabatic cooling and heat transfer from the chamber walls remaining at constant temperature during expansion due to the high heat capacity of the 2 cm thick aluminium walls. After pumping is stopped the gas temperature increases and approaches the wall temperature on a time scale of about five minutes. Volume expansion into an evacuated vessel of 4 m3 volume can additionally be used to sharply increase the supersaturation by up to 20 % within a few seconds. Evaporation of ice phases is forced by controlled adiabatic heating of the chamber gas due to refilling the chamber with dry synthetic air. Water vapour is measured with three independent instruments: The FISH Lyman-a hygrometer of the ICG-1 of Forschungszentrum Jülich [7], the prototype of a novel photoacoustic water vapour sensor (PAS) developed and operated by the University of Szeged, Hungary [8], and a commercial cooled mirror hygrometer M3 from General Eastern. All instruments are operated outside the chamber using the same heated sampling tube. 226 180 Mean Wall Temperature 225 170 224 Expansion and 160 Cooling Period 223 150 222 140 221 130 Pressure (hPa) Temperature (K) 220 120 Pressure 219 Mean Gas 110 Temperature 218 100 0.0 2.5 5.0 7.5 10.0 12.5 15.0 Time (min) Fig. 2 Time profiles of pressure, mean wall temperature, and mean gas temperature during a typical ice activation experiment. 2.3 Detection of ice formation An Argon-Ion laser beam (99% polarized radiation at 488 nm) is conducted into the chamber via an optical fibre which preserves the plane of polarization (Figure 3, left panel). The laser beam and the aperture of the detection optics overlap in the middle of the chamber at a distance of 2 m from the walls, providing about 2 cm3 of scattering volume. The scattered light is split into the parallel and the perpendicular components by a Glan-Taylor prism and then detected by two independent photomultipliers (Figure 3, right panel). Detector optics are mounted at scattering angles of 176∞ and 4∞. Photon counting is employed to achieve high sensitivity and time resolution. The laser source and the detectors can be attenuated by neutral density filters to avoid saturation, and to match the sensitivities of the forward and backward scattering detectors. This setup provides information on the volume, size, and phase of the scattering aerosol. The data set allows for a precise determination of the onset of ice formation and the formation and growth of liquid and solid aerosol particles. AIDA chamber, d = 4 m lens system, pinhole optical fiber detector 2, Glan-Taylor prism o scattering angle 4 Ar-Ion-Laser AIDA PMT 1 Ip dump Data Acquisition PMT 2 o detector 1, scattering angle 176 Is Fig. 3 Overall cross section of the laser scattering device (left) and detail of the detector for measuring backscattered light intensity and depolarisation. Size distributions of aerosol particles before, during, and after periods of ice nucleation are measured with an optical particle spectrometer (PCS2000, Palas) operated below the aerosol chamber (c.f. Figure 1). Residence times in the cold vertical sampling tube are short enough to minimise evaporation of ice particles. Aerosol extinction is measured by long path in situ FTIR spectrometry (Bruker IFS 66v, 254 m folded optical path) in the spectral range 6000 - 800 cm-1 with a resolution of 4.0 cm-1. The FTIR measurements are made with a maximum time resolution of 40 seconds. The extinction spectra provide valuable information about size, chemical composition, and phase of the average particle volume both at equilibrium condition and periods of high cooling and heating rate. 3. HOMOGENEOUS FREEZING OF SUPERCOOLED SULPHURIC ACID PARTICLES Binary sulphuric acid droplets with a mean diameter of about 200 nm are generated at atmospheric pressure outside the aerosol chamber. Aerosol is generated by dispersing a 20 wt% sulphuric acid solution and dried by passing through a glass tube partly filled with a 96 wt% sulphuric acid solution. The dried aerosol is passed into the chamber through a pressure reduction valve and a stainless steel tube. The size distribution of the AIDA aerosol covers the size range of stratospheric sulphuric acid background particles (c.f. Figure 4). LTPLTP DMA Stratosp heric Sulphuric Acid Particles 8000 140140 hPa, 215K 215 K 30.11.199830.11.1998 6000 ogD l dN/d 4000 Data Lognormal Fit 2000 0 10 100 1000 D(nm) Fig. 4 Number size distribution of sulphuric acid particles measured during an AIDA PSC simulation experiment at a temperature of 215 K and a pressure of 140 hPa. Number concentration and size distribution of sulphuric acid aerosol particles are measured with a condensation nuclei counter (CNC3010, TSI) and a differential mobility analyser in combination with a CNC3010. The CNC devices have been modified for operation at pressures from 100 hPa to atmospheric pressure. The modified DMA is operated at the same temperature as the aerosol chamber in order to avoid size change of particles by evaporation processes. Figure 5 shows the result of a freezing experiment with sulphuric acid particles. The expansion was started at t=223 min at a pressure of 180 hPa and a temperature of 202 K. Ice formation occurred after 3 min of pumping at t=226 min, as clearly indicated by the sudden increase of the depolarisation ratio and the number of ice particles. Start pumping Depolarisation No. of ice particles Fig. 5 Depolarisation ratio of backscattered laser light (upper curve) and number of ice particles formed during homogeneous freezing of supercooled sulphuric acid aerosol particles (lower curve). 4. HETEROGENEOUS ICE NUCLEATION OF SOOT PARTICLES The heterogeneous ice nucleation potential of soot particles was investigated at temperatures between Ð62∞C and Ð22∞C. The soot aerosol was taken from a graphite spark generator (GfG1000, Palas). 180 160 140 RHi (%) 120 AIDA (graphitesparc generator soot) DeMott e t al. 1999 (Degussa soot, ~monolayer sulphuric acid) DeMott e t al. 1999 (Degussa soot, multilayer sulphuric acid) Water saturation 100 -70 -60-5 0 -40 -30 -20 Temperature (°C) Fig. 6 Ice activation relative humidities measured for spark generator soot particles (filled squares, present study) and “Degussa” soot coated with monolayer and multilayer sulphuric acid (open and filled diamonds [2]). The relative humidity with respect to ice, RHi, was calculated as function of measured ice frost point and mean gas temperature. Figure 6 shows RHi measured at the onset of ice formation. At higher temperatures liquid water seems to condense on the soot particles before ice activation occurs (immersion freezing). At lower temperatures ice is formed significantly below the liquid water saturation threshold. Figure 6 also depicts results from DeMott et al. [2]. Degussa soot used in that study shows significant lowering of ice onset RHi only for multilayer sulphuric acid coating. The GfG soot used in the AIDA experiments had not been coated with sulphuric acid. 5. SUMMARY AND CONCLUSION In the AIDA experimental facility, ice activation experiments are performed by simulating cycles of ice and water super- and sub-saturations, using the method of expansion-cooling. Relative humidity can dynamically be increased in a controlled manner from ice saturation to values above water saturation within several minutes. First results on homogeneous freezing of super-cooled sulphuric acid particles and heterogeneous ice nucleation of soot prove the AIDA facility to be well suited for IN studies. One of the major advantages of AIDA ice nucleation experiments is the fact that only a minor fraction of the aerosol is lost during expansion cycles. Therefore, the same aerosol sample can be investigated in repeated activation and evaporation cycles. This opens an avenue for future experiments to study the influence of particle ageing effects (e.g. restructuring, coating) on the ice nucleation potential of relevant aerosols. ACKNOWLEDGEMENTS Valuable assistance by Rainer Buschbacher, Meinhard Koyro, Elisabeth Kranz, Georg Scheurig, and Claudia Tromm, is gratefully acknowledged. REFERENCES [1] H.R. Pruppacher, and J.D. Klett, Microphysics of clouds and precipitation (Kluwer Academic Publishers, 1997). [2] P.J. DeMott, Y. Chen, S.M. Kreidenweis, D.C. Rogers, and D.E. Sherman, Geophys. Res. Lett. 26 (1999) 2429. [3] K.S. Carslaw et al., Nature 391 (1998) 675. [4] H. Mehrtens, U. von Zahn, F. Fierli, B. Nardi, and T. Deshler, Geophys. Res. Lett. 26 (1999) 603. [5] J. Schreiner, C. Voigt, A. Kohlmann, F. Arnold, K. Mauersberger, and N. Larsen, Science 283 (1999) 968. [6] A. Waibel et al., Science 283 (1999) 2064. [7] M. Zöger et al., J. Geophys. Res. 104 (1999) 1807. [8] M. Szakáll, Z. Bozóki, M. Krämer, N. Spelten, O. Moehler, and U. Schurath, Environ. Sci. Technol. (submitted). THE ESF SCIENTIFIC NETWORK, SPECIAL M.J. Rycroft1, M. Fuellekrug2, N. B. Crosby3 and A.S. Rodger4 1 Faculty of Computing Sciences and Engineering, De Montfort University, The Gateway, Leicester LE1 9BH, U.K., Email: [email protected] 2 Institute of Meteorology and Geophysics, Feldbergstrasse 47, University of Frankfurt/Main, 60323 Frankfurt/Main, Germany, Email: [email protected] 3 University College London, Mullard Space Science Laboratory, Holmbury St. Mary, Dorking, Surrey RH5 6NT, U.K., Email [email protected] 4 British Antarctic Survey, Madingley Road, Cambridge CB3 0ET, U.K., Email: [email protected] Abstract In May 1999, the European Science Foundation approved funding for the establishment of a new Scientific Network on Space Processes and Electrical Changes Influencing Atmospheric Layers. In this interdisciplinary area, possible links between changes of energetic charged particle fluxes and of weather and climate occurring via electrical processes in the atmosphere are being investigated with a three-pronged approach, with groups being set up on: a) the AC, and DC, global atmospheric electric circuit, i.e. concerned with charges, currents and potential differences in the atmosphere, and its electrical conductivity, b) tropospheric and stratospheric responses to energetic charged particles during well-observed space weather events, and statistical studies of such relationships, c) Schumann resonances of the Earth-ionosphere cavity excited by tropical lightning, and sprites (upward lightning which heats and ionises the upper atmosphere at heights between 70 and 90 km). Some interesting results of collaborative experiments, data analysis, theory and modelling obtained to date are presented. A second phase of the SPECIAL Scientific Network has been approved to run from January 2002 to December 2003. Its purpose is to study the links between solar activity, magnetospheric variability, clouds, thunderstorms and lightning further. 1. BACKGROUND TO SPECIAL The recent Intergovernmental Panel on Climate Change (IPCC 2001) identified the largest unknowns in the climate system as being the effects of a) aerosols and b) the Sun [1]. The way in which the Sun affects the Earth's climate may be through direct processes (e.g., UV radiation) or through indirect processes. Many mechanisms have been suggested; these are reviewed in a recent dedicated issue of Space Science Reviews [2]. The Scientific Network on Space Processes and Electrical Changes Influencing Atmospheric Layers (SPECIAL) is a cross disciplinary scientific programme that addresses two key topics in climate change, aerosols and some effects of the Sun, from a novel perspective. Approved by the European Science Foundation in May 1999, SPECIAL ran until the end of 2000 [3]. Details are given on the website http://www.sgo.fi/SPECIAL for which T. Ulich is responsible. In this study of space weather and Earth's weather phenomena, emphasis is placed on the physics, and physical mechanisms, by which the two may be causally linked. The energy input to different layers of the atmosphere, in situ, from above and from below is considered, as is the energy transfer from one location to another. Whilst atmospheric physics generally discusses fluid dynamics and thermodynamics and the conversion of energy from one form to another on a variety of timescales, in SPECIAL, electrodynamics, and electricity and magnetism, are also directly involved. Thunderstorms and the global atmospheric electric circuit are significant phenomena in SPECIAL. Both linear and non-linear feedback processes may be important: negative feedback stabilises the system, but positive feedback leads to the amplification of an initial perturbation and, perhaps, instability and large amplitude oscillations. Threshold effects, or triggering mechanisms, whereby a small energy input can lead to a large effect, could be significant in the physical linkages here. SPECIAL covers a controversial, but potentially very important, area of research where several processes interact in the complex laboratory which is our atmosphere and which protects us from energetic photons and other forms of radiation from the Sun and the cosmos beyond. In outline, SPECIAL investigates the links between changes in the fluxes of galactic cosmic rays, solar energetic charged particles and precipitating magnetospheric charged particles and changes of the weather and climate occurring via electrical processes in the atmosphere. The aim of SPECIAL is to improve our scientific understanding of such topics within the Sun-Earth scenario. The physical agents responsible for solar variability effects in the Earth's atmosphere may be a) solar ultraviolet radiation, b) charged particles (e.g., cosmic rays, solar particles or magnetospheric particles precipitating into the atmosphere) or c) waves (tides, planetary scale waves, gravity waves, etc.; see, e.g., Arnold and Robinson [4]). Topic a) is not discussed further here because changes of ultraviolet radiation and the stratosphere form the basis of the Stratospheric Processes and their Role in Climate (SPARC) programme. Neither is topic c) specifically considered further. In SPECIAL, the focus is on charged particle effects, from the lowest energies (thermal; eV) in the ionosphere, to magnetospheric (keV), solar proton (MeV) and galactic cosmic ray (GeV) effects in the troposphere [3]. High energy charged particles reach deep into the atmosphere to deposit their energy, or to cause ionisation, there. They can change the electrical conductivity of the atmosphere (which is due to ions); the atmosphere is a good insulator near the Earth's surface, but is highly conducting near the ionosphere, where the very mobile electrons determine the conductivity. 2. SOME RESULTS RELEVANT TO THE SPECIAL PROGRAMME Roederer [5] presented a valuable review of both observational claims and possible mechanisms for solar variability effects on climate. Atmospheric ionisation and the global electric circuit are featured here. In 1959, Ney [6] showed observationally that at solar maximum the tropospheric and stratospheric ionisation was less than at solar minimum. We now know that this 11 year solar cycle effect is due to the enhanced scattering of galactic cosmic rays by irregularities of the interplanetary magnetic field near solar maximum, the mechanism which is also responsible for Forbush decreases. Cosmic ray flux variations may also cause variations of aerosol production in the atmosphere and of cloud droplet nucleation. The change of atmospheric ionisation is sizeable, ranging from ~5% to ~50%: the effect is more marked at higher altitudes and higher latitudes. Bazilevskaya et al. [7] have recently reviewed this topic comprehensively. In the same Golden Jubilee issue of the journal, Rycroft et al. [8] have considered the global atmospheric electric circuit. Their Fig.5 shows the upward current above a typical thundercloud, ~1.3A [9], for about the thousand thunderstorms that are active at any one time, on average. That charges the ionosphere to a potential of ~ +250 kV with respect to the Earth's surface. The return current J ~ 2 pA m-2 flows through the fair weather atmosphere remote from thunderstorms; it is believed that these return currents close through the land/ocean surface and the lowest part of the atmosphere below thunderclouds, but electrified shower clouds could also be important [10]. Following a Forbush decrease, the tropospheric conductivity could decrease by ~10%. If J is unchanged, an increase by ~10% in the fair weather electric field near the Earth's surface, E ~ 130 V m-1, could be expected. However, it is not easy to detect that in a noisy signal due to local meteorological effects. Markson [11] indicated that a 10% change of ground level cosmic radiation was associated with a 10-20% variation of ionospheric potential. By Poisson's equation and Gauss' theorem, which relate the charge density to the electric field distribution through the atmosphere with its conductivity gradient, charged layers may exist near cloud tops. Here, electroscavenging could occur and ice may be produced [12], leading to the formation of clouds. Clouds exert a profound influence on radiative forcing and the radiation balance of the atmosphere, and hence on the weather and climate [1, 12]. Pudovkin and Veretenenko [13] claim that 1 or 2 days after a Forbush decrease in winter the cloudiness at geographic latitudes between 60∞ and 64∞ is about 7% (1 or 2 standard deviations) less than usual. Using the Student's t-test, this is significant at the 98% level. They observe no variations at middle latitudes (~50∞). Marsh and Svensmark [14] find that the total cloud cover over southern hemisphere oceans derived from satellite observations is 1.5% less at solar maximum than at solar minimum when the cosmic ray flux at the mid-latitude Climax station (in Colorado) is ~10% less. Following a Forbush decrease or solar proton event, changes of the atmospheric circulation at 30 hPa (~24 altitude km) and 500 hPa (~5 km) have been discussed by Gabis and Troshichev [15]. King [16] showed that the length of the growing season in Southern Scotland seems to exhibit the 11 year solar cycle variation. If real, this effect is important for the agricultural economies of nations worldwide. Kristjansson and Kristiansen [17] presented a possible causal mechanism between increased solar activity, reduced cosmic ray flux and reduced cloud cover, and - perhaps - a warmer climate. On the other hand, Ney [6] speculated that a reduced cosmic ray flux would lead to decreased atmospheric ionisation and - perhaps - increased thunderstorm activity and a cooler climate. Williams [18] and Price [19] suggested that increased thunderstorm activity is associated with global warming (see also [8, 20]). 3. THE SPECIAL PROGRAMME The planning and carrying out of the work in the framework of SPECIAL have been done in three distinct groups [3]. These three groups, each of a reasonable size, cover: a) the AC, and DC, global atmospheric electric circuit, i.e. concerned with charges, electric currents and potential differences in the atmosphere, and its electrical conductivity. b) tropospheric and stratospheric responses to energetic charged particle fluxes during well observed space weather events, and also statistical studies of such relationships, c) Schumann resonances of the Earth-ionosphere cavity (~8 Hz fundamental) excited by tropical lightning, and also sprites (upward lightning which heats and ionises the upper atmosphere at around 70 to 90 km altitude). The SPECIAL Scientific Network has held two meetings which engendered considerable cross- disciplinary research and discussion. Some new collaborations have been established and the first results are beginning to be published. For example, Neubert et al. [21] have conducted the first European sprite campaign. Rodger et al. [22] have recently shown that sprites could double the night-time ionospheric electron concentration at 90 km altitude. Egorova et al. [23] and Lam and Rodger [24] have been exploring the extent to which Forbush decreases cause changes in the troposphere over the Antarctic in the polar night. Schlegel et al. [25] have demonstrated some solar cycle variations in lightning occurrence over Germany. However, it is equally challenging to explain the lack of a solar cycle effect over Austria. Another relevant recent result is that of Schlegel and Fuellekrug [26], who showed that ionisation at the upper boundary of the Earth-ionosphere cavity increases during solar proton events. This is responsible for an observed increase (~0.1 Hz) of the fundamental resonant frequency (see also [18, 20]), and a decrease of up to 10% in its damping. In the same Golden Jubilee issue mentioned earlier, Barr et al. [27] comprehensively review the topic of ELF/VLF radio wave phenomena, including Schumann resonances; Rodger and Jarvis [28] consider some important topics of ionospheric research, highlighting long term changes and areas of study still required before better long term predictions can be made. A second phase of SPECIAL has recently been approved by the European Science Foundation. The work of the three groups will thus continue until the end of December 2003, the objectives being to investigate the links between solar activity, magnetospheric variability, clouds, thunderstorms and lightning further. This interdisciplinary arena is especially challenging. 4. REFERENCES [1] Houghton J.T., Ding Y., Griggs D.J., Noguer M., van der Linden P.J. and Xiaosu D. (ed.), Climate change 2001: The scientific basis, Cambridge University Press, pp.944, 2001. [2] Friis-Christensen E., Frohlich C., Haigh J.D., Schlusser M. and von Steiger R. (ed.), Solar Variability and Climate, Space Science Reviews, Vol. 94, 1-427, 2000. [3] Crosby N.B. and Rycroft M.J., SPECIAL: an interdisciplinary ESF network on space weather and the Earth's weather, Proc. 1st Solar and Space Weather Euroconference, 'The Solar Cycle and Terrestrial Climate', ESA SP-463, 219-221, 2000. [4] Arnold N.F. and Robinson T.R., Solar cycle changes to planetary wave propagation and their influence on the middle atmosphere circulation, Ann. Geophysicae, Vol. 16, 69-76, 1998. [5] Roederer J.G., Solar variability effects on climate, ESF Workshop, Solar Output and Climate during the Holocene, Bologna, Italy, April, 1993. [6] Ney E.P., Cosmic radiation and the weather, Nature, Vol. 183, 451-452, 1959. [7] Bazilevskaya G.A., Krainev M.B. and Makhmutov V.S., Effects of cosmic rays on the Earth's environment, J. Atmos. Sol.-Terr. Phys., Vol. 62, 1577-1586, 2000. [8] Rycroft M.J., Israelsson S. and Price C., The global atmospheric electric circuit, solar activity and climate change, J. Atmos. Sol.-Terr. Phys., Vol. 62, 1563-1576, 2000. [9] Stergis G.G., Rein G.C. and Kangas T., Electric field measurements above thunderstorms, J. Atmos. Terr. Phys., Vol. 11, 83-90, 1957. [10] Markson R., Private communication, 2001. [11] Markson R., Modulation of the Earth's electric field by cosmic radiation, Nature, Vol. 291, 304-308, 1981. [12] Tinsley B.A., Influence of solar wind on the global electric circuit, and inferred effects on cloud microphysics, temperature, and dynamics in the troposphere, in [2], Solar Variability and Climate (ed. Friis-Christensen E., Frohlich C., Haigh J.D., Schlusser M. and von Steiger R.), Kluwer Academic Publishers, Dordrecht, The Netherlands, 231-258, 2000. [13] Pudovkin M.I. and Veretenenko S.V., Variations of the cosmic rays as one of the possible links between the solar activity and the lower atmosphere, Adv. Space Res., Vol. 17, (11)161- (11)164, 1996. [14] Marsh N. and Svensmark H., Cosmic rays, clouds and climate, in [2], Space Science Reviews, Vol. 94, 215-230, 2000. [15] Gabis I.P. and Troshichev O.A., Influence of short-term changes in solar activity on baric field perturbations in the stratosphere and troposphere, J. Atmos. Sol.-Terr. Phys., Vol. 62, 725-735, 2000. [16] King J.W., Solar radiation changes and the weather, Nature, Vol. 245, 443-446, 1973. [17] Kristjannson J.E. and Kristiansen, J., Is there a cosmic ray signal in recent variations in global cloudiness and cloud radiative forcing? J. Geophys. Res., Vol. 105, 11851-11863, 2000. [18] Williams E.R., The Schumann resonance: a global thermometer, Science, Vol. 256, 1184- 1187, 1992. [19] Price C., Global surface temperatures and the atmospheric electrical circuit, Geophys. Res. Lett., Vol. 20, 1363-1366, 1993. [20] Price C., Evidence for a link between global lightning activity and upper tropospheric water vapour, Nature, Vol. 406, 290-293, 2000. [21] Neubert T., Allin T.H., Stenback-Nielsen H. and Blanc E., Sprites over Europe, Geophys. Res. Lett., Vol. 28, 3585-3588, 2001. [22] Rodger C.J., Cho M., Clilverd M.A. and Rycroft M.J., Lower ionospheric modification by lightning-EMP: simulation of the night ionosphere over the United States, Geophys. Res. Lett.,Vol. 28, 199-202, 2001. [23] Egorova L.V., Vovk V.Y. and Troshichev O.A., Influence of variations of the cosmic rays on atmospheric pressure and temperature in the Southern geomagnetic pole region, J. Atmos. Sol.-Terr. Phys., Vol. 62, 955-966, 2000. [24] Lam M.M. and Rodger A.S., The effects of Forbush decreases on tropospheric parameters over the Antarctic, J. Atmos. Sol.-Terr. Phys. (in press). [25] Schlegel K., Diendorfer D., Thern S. and Schmidt S., Thunderstorms, lightning and solar activity - Middle Europe, J. Atmos. Sol.-Terr. Phys., Vol. 63, 1715-1728, 2001. [26] Schlegel K. and Fuellekrug M., Schumann resonance parameter changes during high-energy particle precipitation, J. Geophys. Res., Vol. 104, 10111-10118, 1999. [27] Barr R., Jones D.L. and Rodger C.J., ELF and VLF radio waves, J. Atmos. Sol.-Terr.Phys., Vol. 62, 1689-1718, 2000. [28] Rodger A.S. and Jarvis M.J., Ionospheric research 50 years ago, today and tomorrow, J. Atmos. Sol-Terr. Phys., Vol. 62, 1629-1645, 2000. CLOUD: A PARTICLE BEAM FACILITY TO INVESTIGATE THE INFLUENCE OF COSMIC RAYS ON CLOUDS Jasper Kirkby CERN, Geneva, Switzerland Abstract Palaeoclimatic data provide extensive evidence for solar forcing of the cli- mate during the Holocene1 and the last ice age, but the underlying mecha- nism remains a mystery. However recent observations suggest that cosmic rays may play a key role. Satellite data have revealed a surprising correla- tion between cosmic ray intensity and the fraction of the Earth covered by low clouds [1, 2]. Since the cosmic ray intensity is modulated by the solar wind, this may be an important clue to the long-sought mechanism for solar-climate variability. In order to test whether cosmic rays and clouds are causally linked and, if so, to understand the microphysical mechanisms, a novel experiment known as CLOUD2 has been proposed [3]–[5]. CLOUD proposes to inves- tigate ion-aerosol-cloud microphysics under controlled laboratory conditions using a beam from a particle accelerator, which provides a precisely adjustable and measurable artificial source of cosmic rays. The heart of the experiment is a precision cloud chamber that recreates cloud conditions throughout the atmosphere. 1INTRODUCTION That there is a causal connection between the observed variations in the forces of the Sun, the terrestrial magnetic field, and the meteorological elements has been the conclusion of every research into this subject for the past 50 years. The elucidation of exactly what the connection is and the scientific proof of it is to be classed among the most difficult problems presented in terrestrial physics. The evidence adduced in favor of this conclusion is on the whole of a cumulative kind, since the direct sequence of cause and effect is so far masked in the complex interaction of the many delicate forces in operation as to render its immediate measurement quite impossible in the present state of science. F.H. Bigelow US Dept. Agriculture Weather Bureau Bulletin No.21, 1898 This quotation [6] is from an article written over a century ago and yet it could be taken almost wholly from a contemporary paper. The observation that warm weather seems to coincide with high sunspot counts and cool weather with low sunspot counts was made as long ago as two hundred years by the astronomer William Herschel [7] who noticed that the price of wheat in England was lower when there were many sunspots, and higher when there were few. The most well-known example of a solar-climate effect is known as the Maunder Minimum [8], the period between 1645 and 1715— which ironically almost exactly coincides with the reign of Louis XIV, le Roi Soleil, 1643–1715—during which there was an almost complete absence of sunspots (Fig. 1). This marked the most pronounced of several prolonged cold spells in the period between about 1450 and 1890 which are collectively known as the Little Ice Age. During this period the River Thames in London regularly froze across and fairs 1The Holocene is the present interglacial period—the previous 11.5 kyr since the end of the last ice age. 2CLOUD is an acronym for Cosmics Leaving OUtdoor Droplets. 300 300 200 200 100 100 Sunspot number Maunder Minimum 0 0 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 Year 300 300 200 200 Dalton Minimum 100 100 Sunspot number 0 0 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 Year Fig. 1: Variation of the sunspot number from 1610 to 2001. The record starts 3 years after the invention of the telescope by Lippershey in Holland. The Maunder and Dalton Minima are two pronounced cold spells in the period between about 1450 and 1890 which are collectively known as the Little Ice Age. complete with swings, sideshows and food stalls were a standard winter feature. Numerous studies of palaeoclimatic proxies have both confirmed that the Little Ice Age was a global phenomena and shown that it was but one of around 10 occasions during the Holocene when the Sun entered a grand minimum for centennial-scale periods and influenced the Earth’s climate (§3). During the Holocene there have also been a similar number of extended periods of high solar activity, amongst which is the second half of the 20th century. However, despite the evidence, solar variability remains controversial as a source of climate change since no causal mechanism has been established to link the two phenomena. The most obvi- ous mechanism to suspect is a variation of the solar irradiance. Precision satellite measurements of the solar irradiance have indeed revealed a small variation of about 0.1% over the solar cycle [9] (§4.2). Together with observations of cyclic stars similar to the Sun, this has led to estimates of somewhat larger long-term variations of the solar irradiance, but these nevertheless appear to be too small to account for the observed climate variability. For example, it is estimated that the solar irradiance, I,was weaker by 3.3 Wm−2 (∆I/I =2.4 · 10−3) during the Maunder Minimum [10], when globally-averaged tempera- tures were cooler by about 0.5–1K, after subtracting the estimated anthropogenic contributions during the last century. The relative temperature change is then ∆T/T =(0.5 − 1.0)/288 = (1.7 − 3.5) · 10−3. This suggests that the Earth’s temperature sensitivity, ∆T/T ∆I/I. Since a simple black body would respond as ∆T/T =∆I/4I, this implies either that the Earth has a high sensitivity to irradiance changes or that other mechanisms exist that amplify the solar variations, or both. Indeed, the response of climate is complex and involves more than simple radiative heating and cooling (§3.3.2). Moreover there is in fact no direct evidence that the irradiance of the Sun is varying beyond the 0.1% solar cycle variations (which can be quantitatively well-explained by sunspot darkening and facular brightening of the pho- tosphere). So the magnitude of the long-term change in solar irradiance—if any—is speculative. The physical mechanism or mechanisms for solar-climate variability therefore remain a mystery. However, the recent observation of correlations between the galactic cosmic ray (GCR) intensity and the fraction of Earth covered by low clouds [1, 2] (§2.1) may provide an important clue. Clouds cover a large fraction of the Earth’s surface—a global annual mean of about 65%—and exert a strong net cooling effect of about 30 Wm−2,solong-term variations of only a few per cent could have a significant effect on the Earth’s climate. Since the GCR intensity is modulated by the solar wind, a GCR-cloud link could provide a sufficient amplifying mechanism for solar-climate variability. This would constitute a new solar indirect contribution to climate change, in addition to the direct contribution from irradiance changes (§3.3.5). If a causal connection between GCR intensity and low cloud cover were to be confirmed, it could have profound consequences for our understanding of the solar contributions to the current global warm- ing. During the 20th century the Sun’s magnetic activity increased dramatically and the solar wind more than doubled in strength [11] (§3.3.1). As a consequence, the mean GCR intensity on Earth diminished by about 15%. The implied reduction in low cloud cover by about 1.3% absolute could have given rise to a radiative forcing of about +0.8 Wm−2 (3.3 · 10−3), which is comparable to the estimated total anthropogenic forcing of about +1.3 Wm−2 (§3.3.5). We can look further back in time for evidence of solar forcing3 of the climate. Detailed records of the magnetic variability of the Sun are preserved in the light radio isotope archives, notably the 14C content of tree-rings (for about the last 10 kyr) and the 10Be concentrations in ice-cores from Greenland and Antarctica (about 250 kyr) (§3.1.1). The light radio isotopes are produced by GCRs interacting with nitrogen, oxygen and argon nuclei in the atmosphere, and so they are a direct measure the prevailing GCR intensity. These records show the Sun to be a variable star, with both quasi-cyclic activity (11, 88, 208 yr...) and also periods of ‘grand-minima’ occurring on millennial-scale intervals. Diverse palaeo- climatic records have also shown that the Earth’s climate was not stable in the past and that large changes have occurred naturally. Comparisons between the solar and palaeoclimatic records reveal unmistakable evidence for a solar forcing of the climate (§3.2–§3.3). However, in the absence of an established physical mechanism, even the evidence for solar-climate correlations accumulated in studies over the last two hundred years has not proved cause and effect. But now—and perhaps for the first time—we have a definite hypothesis for the mechanism that can be tested experimentally, namely: are cosmic rays affecting cloud formation? Since the energy flux from cosmic rays is tiny—about the same as that from starlight, and only a few parts per billion compared with the solar irradiance—a strong amplification mechanism would be required, i.e. some microphysical property or properties of clouds must be very sensitive to the ionisation or radicals produced by GCRs. Although a frequent criticism of the GCR-cloud hypothesis has been the absence of any microphysical mechanism, there are in fact several candidates, associated with aerosols, ice particles and cloud electricity (§4). However since none of these mechanisms is firmly established, they must tested experimentally. How can this best be done? In the atmosphere it is hard to establish cause and effect since it is difficult to measure all the variables and essentially impossible to adjust them. For these reasons the CLOUD experiment [3]–[5] (§5) proposes to perform the necessary measurements under controlled con- ditions in the laboratory. CLOUD plans to use a particle beam from an accelerator to provide a precisely controllable source of relativistic ionising radiation that closely duplicates cosmic rays in the atmosphere (§5.4). Processes can be studied with beams of varying intensity around naturally-occurring levels, and also with no beam present. The beam will pass through an expansion cloud chamber (§5.3.1) and a re- actor chamber where the atmosphere is duplicated by moist air charged with selected aerosols and trace condensable vapours. The cloud chamber dynamically simulates the thermodynamic conditions, electric 3A climate forcing is a perturbation of the Earth’s radiative energy balance, with the convention that a positive forcing leads to a warming, and a negative forcing to a cooling. fields and water vapour supersaturations within clouds throughout the troposphere and stratosphere. As well as in situ analysis of the cloud chamber contents, samples are extracted and sent to an array of exter- nal detectors and mass spectrometers where the physical and chemical characteristics of the aerosols and trace gases are analysed during beam exposure. Where beam effects are found, the experiment will eval- uate their significance in the atmosphere by incorporating them into aerosol and cloud models, and by examining the sensitivity of clouds under atmospheric conditions to variations of the GCR intensity in the presence of other sources of natural variability. A close exchange is foreseen between CLOUD and the related field experiments so that, on the one hand, the laboratory results can be applied in the atmosphere and, on the other hand, new field work can help to shape the CLOUD experimental programme. CLOUD is designed as a flexible ‘general-purpose’ detector, for which a wide range of experi- ments on ion-aerosol-cloud interactions is envisaged over several years (§5.5). Flexibility is required because this field is relatively unexplored but likely to develop rapidly in the coming years—and it is impossible to predict where these future experimental and theoretical developments may lead. For these reasons it is more appropriate to consider CLOUD as a facility than a one-off experiment. As well as its primary goal of investigating the effect of cosmic rays on clouds, the CLOUD facility will provide valu- able experimental data on a broad range of important related aerosol and cloud properties, such as the optical reflectivities from liquid and ice clouds, and the dynamics of the activation of cloud condensation nuclei (CCN) into droplets. 2 OBSERVATIONS OF SOLAR-CLOUD VARIABILITY 2.1 Experimental observations Although clouds have been routinely monitored from ground stations for more than a century, it was only over the last 20 years that global measurements became available from satellites. In 1997 Svensmark and Friis-Christensen reported a surprising correlation between global cloud cover and the GCR intensity [1]. Following their discovery, several papers pointed out important limitations in the satellite cloud data and its analysis (see, for example, refs. [12]–[14]). Among the concerns were the use of a composite of several independent satellite datasets with limited time coverage and with inter-calibration uncertainties; aspatial coverage limited to oceans and excluding the tropics and polar regions; a limited temporal coverage (mostly daytime only); and the absence of any indication of which type of cloud is affected (an increase in high clouds would result in a warming whereas an increase in low clouds causes a cooling). Afrequent—but misplaced—criticism has also been the lack of any physical mechanism connecting cosmic rays and cloud cover (§4). These limitations have largely been addressed with the recent release of the ISCCP-D2 cloud dataset [15] and its subsequent analysis [2]. The new cloud data (Fig. 2) comprise a single unified dataset over the period July 1983 to September 1994 and provide complete global coverage, day and night (at 10–12 µmIRwavelengths). As well as cloud frequency, the cloud-top temperatures and pressures are also determined. The temperatures and pressures are obtained by assuming an opaque cloud, i.e. an emissivity =1, and adjusting the cloud’s pressure level (effectively the cloud-top altitude) in the model until the reconstructed outgoing IR flux matches that observed. The clouds are classified into 3 altitude ranges according to the pressure at their top surface: low, >680 hPa (approximately <3.2 km); middle, 680–440 hPa (3.2–6.5 km); and high, <440 hPa (>6.5 km). The new cloud data indicate the presence of a solar modulation in the fraction of low clouds—but none for clouds at higher altitudes (Fig. 2). This establishes the sign of the GCR-cloud correlation: a higher GCR intensity is associated with increased low clouds and therefore with a cooler temperature (§2.2). The global distribution of the correlation of GCR intensity and low cloud fraction is shown in Fig. 3a) [2]. The fraction of the Earth’s surface with a correlation coefficient above 0.6 is 14.2%. The regions of high correlation appear to be rather uniformly distributed—although there appears to be some preference for the oceans, where aerosol concentrations are generally lower than over land (§4.4.2). A es 2 n 1].TeICPD lu aaaeotie tifard(R aeegh (10–12 wavelengths (IR) infra-red at obtained are data cloud ISCCP-D2 The [16]). and [2] refs. h engoa lu rcinoe hspro o ih ideadlwI lusi 35,1.% n 28.0% and 19.9%, 13.5%, is [15]. clouds dataset IR IR low ISCCP-D2 indicated. and the also from middle are obtained line; high, scale) are for (dashed inverted measurements period note intensity cloud this panel; ray The over lowest respectively. cosmic fraction in of cloud line period, global (dotted this mean irradiance over The solar and variations, cutoff) The rigidity GeV/c night. 13 and day coverage, global complete pressure top 2: Fig. otl envle o h lblaslt aitoso nr e I)codcvrg o )hg (cloud- high a) for coverage cloud (IR) red infra of variations absolute global the for values mean Monthly < 4 P) )mdl 4060ha,adc o ( low c) and hPa), (440–680 middle b) hPa), 440 Change in cloud fraction (%) Change in cloud fraction (%) Change in cloud fraction (%) -1 -1 -1 0 1 2 0 1 2 0 1 2 9519 1995 1990 1985 1995 1990 1985 1995 1990 1985 a) Highclouds c) Low clouds b) Middleclouds cosmic rays (>13GeV) solar irradiance clouds cosmic rays (>13GeV) clouds cosmic rays (>13GeV) clouds ear a Ye r a Ye r a Ye > 8 P)cod sldlns aatdfrom (adapted lines) (solid clouds hPa) 680 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 Change in cosmic rays (%) Change in cosmic rays (%) Change in cosmic rays (%) 0.10 0.05 0 -0.05 -0.10 Change in solar irradiance (%) µ )adhave and m) Fig. 3: Global maps of the correlation between cosmic ray intensity and a) low IR cloud fraction and b) low IR cloud-top temperature [2]. The low IR cloud fractions are calculated as in Fig. 2c), while the low cloud-top temperatures are obtained from the ISCCP-D2 IR model. White pixels indicate regions with either no data or an incomplete monthly time series. The correlation coefficients are calculated from the 12-month running mean at each grid point. Fractions of the Earth with a correlation coefficient ≥ 0.6 are a) 14.2%, and b) 29.6%, respectively. The probability of obtaining a correlation coefficient ≥ 0.6 from a random signal is < 0.01% per pixel. few regions, such as North America, show a negative correlation. The lower map (Fig. 3b) shows the correlation of low IR cloud-top temperature and cosmic ray intensity. It shows a strong and continuous band of high correlation (>0.6) extending throughout the tropics, covering 29.6% of the globe. This is a counter-intuitive result since the solar modulation of the GCR intensity is a minimum near the geomagnetic equator. Nevertheless there is a finite GCR modulation of about 5% peak-to-peak at the equator. The reason for the different global distributions of high GCR correlation in Figs. 3a) and 3b) is not known, although we note that these are two distinct cloud properties and so, in principle, they may be affected differently. Cloud frequency is a measure of cloud lifetime whereas cloud-top temperature measures the altitude at its upper boundary. The striking band of high correlation seen in the cloud-top temperatures over essentially the entire tropics may indicate an influence of GCRs on the convective ac- tivity of the Inter Tropical Convergence Zone (ITCZ)—the boundary between the northern and southern Hadley cells, where the Earth’s most intense convective transport of water vapour into the upper tropo- sphere occurs. There is, in fact, some palaeoclimatic data to support a possible link between the ITCZ and solar activity (§3.2.4). The cloud data of Fig. 2 span only a single solar cycle and one may speculate how the correlation will develop in future. There have been numerous previous observations of solar cycle effects on the Earth’s climate that have persisted for some decades and then apparently disappeared [17]. A notable example was the observation in 1923 [18] that the levels of the central African Lakes Victoria and Albert were highly correlated with solar activity (0.87 correlation coefficient) over the previous two solar cycles (1896–1922). The correlation broke down around this time, as did a number of other solar-climate relationships elsewhere. This may suggest the association was accidental. However—and perhaps more likely in view of the coincidental termination of several solar-climate observations—it may reflect the complexity of the Earth’s climate, in which many factors are important and they interact in a complex way. The climate may have “stable” states where the conditions are favourable for solar forcing, and a correlation may persist for some decades. Then, at other times, the conditions are unfavourable and the correlations disappear. Finally we note that there is some indication that the reflectivity (cloudiness) of Neptune may correlate with solar activity [19]. Measurements of the reflectivity at 472 nm and 551 nm from 1972 to 2000 show a 10% overall increase of brightness together with an apparent 2% residual solar modulation that is anti-correlated with solar activity over the three solar cycles spanned by the data. Since Voyager measured Neptune’s magnetic field to be small—less than 1 Gauss—the cosmic ray intensity on Neptune is expected to be modulated by the solar wind. The conditions on Neptune are of course completely different than those on Earth—the solar irradiance is 0.1% of Earth’s, and the clouds are probably liquid methane—so it is not possible to draw any conclusions about the Earth’s climate from this observation. 6 4 2 0 -2 Normalised (%) variation -4 low clouds low clouds - corrected cosmic rays -6 1985 1990 1995 2000 Ye a r Fig. 4: Recent extension of the ISCCP-D2 low IR cloud data up to December 1998 (solid line). The broken line shows the ISCCP data after correcting by its difference with the SSMI cloud data after January 1994. The variations of cosmic ray intensity at Huancayo (13 GeV/c rigidity cutoff) are indicated by the dotted line. All curves have been normalised to their respective mean and variance over the period July 1987 - June 1990 [20]. 2.1.1 ISCCP extension Very recently, an extension of the ISCCP-D2 cloud data has been released for the period from January 1994 to December 1998. This shows a weakening of the correlation between low cloud amount and low cloud top temperature with cosmic rays after 1994 (Fig. 4) [20]. However comparison with independent cloud data from the SSMI instrument4 shows a good agreement with ISCCP low cloud amount until 1994, after which the two measurements seem to diverge. This suggests the possible presence of long term drifts in at least one of the satellite data sets. Indeed it is generally accepted that the long term stability and calibration of multi-satellite cloud detectors, such as ISCCP or SSMI, is challenging. Although it 4The SSMI (Special Sensor Microwave Imager) instrument is part of the DMSP (Defence Meteorological Satellite Program) satellites. is not possible to resolve these discrepancies at present, an estimate of the uncertainties in the drift can be evaluated from the difference in the long term trends of the ISCCP and SSMI data. One limit is to correct the ISCCP cloud data after 1990 in direction and magnitude by its difference with the SSMI data, assuming the two datasets agree over the period 1987–1990. This is shown in Fig. 4 and indicates a good correlation between low cloud amount and cosmic rays over the full period of available cloud data (1983-1998). The uncertainties in cloud amount for the period after 1994 appear to be too large at present to draw any conclusion on either the absence or presence of a correlation with the GCR intensity. 2.2 Cloud radiative forcing The observed variation of low cloud cover over the solar cycle of about 1.7% absolute corresponds to 6.0% relative. Since measurements by the Earth Radiation Budget Experiment (ERBE) indicate that low clouds contribute a global annual mean radiative forcing of about -17 Wm−2 (Table 1) this implies the cloud modulation corresponds to about +1.0 Wm−2,solar minimum-to-maximum. This is about a factor 5 larger than the solar cycle irradiance forcing at the Earth’s surface (0.2 Wm−2; §3.3.3), and in phase. Table 1: Global annual mean forcing due to various types of clouds, from the Earth Radiation Budget Experiment (ERBE) [21]. Parameter High clouds Middle clouds Low clouds Total Thin Thick Thin Thick All Global fraction (%) 10.1 8.6 10.7 7.3 26.6 63.3 Forcing (relative to clear sky): Albedo (SW radiation) (Wm−2) -4.1 -15.6 -3.7 -9.9 -20.2 -53.5 Outgoing LW radiation (Wm−2) 6.5 8.6 4.8 2.4 3.5 25.8 Net forcing (Wm−2) 2.4 -7.0 1.1 -7.5 -16.7 -27.7 3RECORD OF SOLAR-GCR-CLIMATE CHANGE 3.1 Solar and palaeoclimatic records 3.1.1 Solar and GCR records The flux of galactic cosmic rays reaching the Earth’s atmosphere is modulated by variations of the helio- spheric magnetic field and of the Earth’s geomagnetic field. During times of high solar activity (sunspot maximum) there is an increase of the open magnetic flux and of the magnetic irregularities carried out into the heliosphere by the solar wind. These magnetic fields scatter the low-energy component of the incoming GCRs (below a few tens of GeV) and, in consequence, the flux reaching Earth is reduced. The global average modulation of the GCR intensity over the solar cycle is about 15%, but larger variations occur on longer timescales. During a reversal of the Earth’s dipole field, for example, it is estimated that the global GCR rate is enhanced by about a factor 2.5 relative to the present values. An exquisite record of the variations of GCR intensity over the past 250 millennia is preserved in the light radio isotope records in ice cores [22]. These provide an essentially direct measurement of the prevailing GCR intensity and hence are a direct indication of variations of the solar magnetic activity. They are also frequently used as a proxy for putative changes of solar irradiance, although there exists no direct evidence for long-term variations of the solar irradiance. The light radio-isotopes are produced in spallation interactions of GCRs on nitrogen, oxygen and argon nuclei in the atmosphere. The two radioisotopes with the highest production rates are 14C(half life = 5730±40 yr and global mean production rate ∼2.0 atoms cm−2s−1) and 10Be (1.5 Myr; ∼1.8×10−2 atoms cm−2s−1). The third most abundant isotope is 36Cl, which is produced from GCR interactions with Ar nuclei (300 kyr; ∼1.9×10−3 atoms cm−2s−1). 14 14 The Cisrapidly oxidised to CO2. The turnover time of CO2 in the atmosphere is quite short— about 4 years—mostly by absorption in the oceans and assimilation in living plants. However, because of recirculation between the oceans and the atmosphere, changes in the 14Cfraction on timescales less than afew decades are smoothed out. Plant material originally contains the prevailing atmospheric fraction of 14Cand, subsequently, since the material is not recycled into the atmosphere, the fraction decreases with the characteristic half life of 14C. By analysing the 14C content in the rings of long-lived trees such as the California bristlecone pine, a continuous yearly record of GCR intensity over the last 10–15 kyr has been assembled. In the case of 10Be, after production it rapidly attaches to aerosols and follows the motion of the surrounding air masses. Since the production of 10Be follows the intensity profile of the cosmic ray hadronic showers, about 2/3 is produced in the stratosphere and 1/3 in the troposphere, globally averaged. Due to the tropopause barrier, aerosols in the stratosphere take about 1–2 years to settle on the Earth’s surface, whereas the mean residence time in the troposphere is only about a week. If the sedimentation occurs in the form of snow in a permanently frozen and stable region such as Greenland or Antarctica then the subsequent compacted ice preserves a temporal record in layers according to their depth. The measured variations of the light radionuclides are the product of two processes: 1) the produc- tion rate and 2) system effects i.e. transport, precipitation and exchange processes between the different reservoirs. Since the system effects are quite different for 14C and 10Be, it has been possible to reliably determine the production rates, and hence the GCR intensities. An advantage of 14Cisthat it is well- mixed before storage in the tree-ring archives. In contrast, the short residence time of the tropospheric 10Be fraction means that the measured concentrations are subject to possible variations of precipitation rate and of wind directions in carrying the radionuclide from where it is produced to where it is sedi- mented as snow. However, by using a minimum sample period of 1 or 2 years, the effects of variations of transport direction and efficiency are minimised. Since 10Be is not recycled into the atmosphere, it has the important advantages of being able to record relatively short-term changes in the GCR intensity, and afreedom from uncertainties due to variations in the recycling processes. 3.1.2 Palaeoclimatic records Many ingenious proxies have been developed to reconstruct the climate prior to the last two centuries, for which instrumental records are available. Cultural records over about the last millennium are an important source since humans are sensitive to climate change, especially when prolonged drought, cold or flooding is involved. These sources include documents recording the dates when the first cherry blossoms appeared each spring in China, as well as records of the grape harvests in Europe. Other records (with their approximate time span BP 5 in parentheses) are corals (400 yr), tree rings (10 kyr), mosses (10 kyr), pollen (1 Myr), ice cores (250 kyr), ocean sediments (>1 Myr) and geomorphology (3 Byr). Ice cores are an especially valuable record of past climate [22]. As well as the solar-GCR record described above, the trapped gases preserve the atmospheric composition at earlier times, layer thick- nesses measure precipitation rate, dust content measures wind speed and volcanic activity, sulphate measures sulphuric acid content of the atmosphere (volcanic and planktonic activity) and, of particu- 18 lar importance, H2 O measures past temperatures. The physical basis for proxy temperature measurements from the stable 18O isotope is that the 18 16 vapour pressure of H2 Oislower than that of H2 O. Evaporation from the oceans thus produces water vapour that is 18O-depleted (by about 1% relative); conversely, the remaining water is enriched in 18O. 5BP signifies before present, where ‘present’ means 1950. 18 During condensation, the lower vapour pressure of the H2 O leads to preferential condensation, and so the water vapour becomes progressively more 18O-depleted as it travels poleward. Because condensation is the result of cooling, the greater the fall in temperature, the lower is the heavy isotope concentration. Isotope concentration in the condensate is thus a function of the temperature at which condensation 18 16 occurs. The relative proportion of O and Oinanice core sample, Rs,isexpressed in terms of its fractional deviation, δ18O, from a standard value, 18 δ O(per mil) = ((Rs/RSMOW) − 1) · 1000 (1) 18 16 where R =[ O] / [ O], for which the Standard Mean Ocean Water (SMOW) value is RSMOW = 2.0052 · 10−3.Itisfound that a decrease of δ18Oby1per mil corresponds to a temperature decrease at the site of precipitation of between 1.5 K (polar regions) and 1.7 K (mid-latitudes). In addition the δ18O value of sea sediments provides a measure of the global volume of water locked up in (18O-depleted) ice sheets, since high ice volumes leave the oceans enriched in 18O. Deviations of other isotopes are defined in a similar way as δ18O. In the case of a radioisotope like 14C, the final deviation is expressed as ∆14Ctosignify correction of the measured value, δ14C, for radioactive decay and for isotopic (δ13C) fractionation (§3.2.7). Many studies of past climate change show a correlation with changes of the GCR intensity and solar activity [17, 23]. In these cases a colder climate is found to correlate with low solar activity (high GCR intensity) and, conversely, a warmer climate correlates with high solar activity (low GCR intensity). The correlation with rainfall may be in either direction, depending on the region studied and the prevailing climatic conditions. In the remainder of this section we first present some examples of possible solar-influences on climate change during the late glacial and Holocene periods, and then close with a discussion of the solar contribution to the current global warming. 3.2 Solar-GCR-climate change during the late glacial and Holocene 3.2.1 The Younger Dryas (12,700Ð11,550 yr BP) The Younger Dryas cold event (so called because it was marked by the spread of an Alpine flower known as Dryas octopetala) occurred between 12,700 and 11,550 years ago (Fig. 5a). For 3,000 years before the start of the Younger Dryas, the Earth had been gradually warming up after the end of the last ice age, but then the climate abruptly swung back into ice age conditions. During this warming period the first humans had entered the American continent by walking across the Bering land-bridge into Alaska. A settlement excavated from a peat bog at Monte Verde in southern Chile shows that they had rapidly migrated far south. However, around the start of the Younger Dryas, the Monte Verde water table rose and their settlement was flooded and abandoned. The cold Younger Dryas climate continued for about a thousand years before it abruptly switched back to warm conditions, marking the start of the Holocene. The temperature transitions were very rapid; the end of the Younger Dryas saw an increase of polar temperatures by about 15◦C, with half that transition occurring in less than 15 years [24]. It is thought that this event was driven by changes of the ocean circulation. At present, northern maritime Europe is warmed by heat carried polewards by the Gulf Stream. When the warm water meets cold polar air in the North Atlantic, heat is released to the atmosphere and the water cools and sinks. This is reinforced by the increases in salinity, and therefore density, due to evaporation and to the formation of sea ice in the Arctic regions. The descending current is called the North Atlantic Deep Water (NADW). It flows southward through the western Atlantic where it joins the Southern Ocean Deep Water descending off the edges of Antarctica and flowing in an easterly direction. The deep water continues round South Africa and then into the Indian and northern Pacific Oceans, where it surfaces. The North Atlantic is warmer than the North Pacific. The increased evaporation therefore serves to increase salinity relative to the North Pacific, and it is this salinity gradient that is thought to drive the global thermohaline ocean circulation. a) -32 Younger Dryas -34 -36 O 18 δ -38 -40 -42 b) 60 measured ∆14C 40 20 C 14 0 ∆ -20 -40 derived ∆14C from 10Be 10 11 12 13 14 15 Age (kyr BP) Fig. 5: The Younger Dryas cold event: a) the δ18Ovariation over the period 15–9.4 kyr BP, as measured in the Greenland GRIP ice core, and b) the measured ∆14Cvariation over this period (heavy curve) and the derived ∆14Cfrom 10Be ice core data, taking into account some changes in deep water formation (light curve) [28]. This ‘heat conveyor belt’ is quite sensitive to climatic conditions—especially to the amount of fresh water entering the North Atlantic. During a glacial period, the formation of the NADW is thought to be much reduced or even shut down. At these times, the Arctic ice sheet extends much further south into the North Atlantic, pushing the position of the polar front southwards. Cooler sea surface temperatures reduce evaporation and therefore salinity, further weakening the thermohaline circulation. It has been suggested that the onset of the Younger Dryas was triggered by a sudden shutdown of NADW formation and therefore of the global thermohaline ocean circulation. Various causes have been proposed such as the presence of a large amount of fresh water from melting icebergs, as well as the melting and abrupt opening of the St. Lawrence waterway into the North Atlantic, diverting the drainage of fresh water over avast region of North America away from the south and towards the north-east. However shutdown of the NADW alone is considered insufficient to initiate global temperature changes and ice sheet development [25]. Other mechanisms would need to be invoked—and on a global scale since the Younger Dryas is also registered in the tropics and the Southern Hemisphere. Recent measurements of sea-surface temperatures (SSTs) in the mid southern latitudes over the period 40–10 kyr BP [26, 27] support the picture that North Atlantic thermohaline circulation is insufficient to drive the observed climate changes. Interestingly, during the Younger Dryas a large increase occurred of atmospheric 14C. It has been 14 argued that this was due to the reduced circulation of ( C-depleted) CO2 from the oceans as the ice sheets advanced. However the increase of 14C occurs abruptly at the start of the Younger Dryas and seems to be too sharp to be caused by changes of ocean circulation alone. Indeed a recent comparison with the 10Be record during this period has concluded that the largest part of the increase of 14C during the Younger Dryas can be attributed to a change in the production rate, i.e. to an increase of the GCR intensity (Fig. 5b) [28]. This suggests that solar forcing may have triggered and helped to sustain the Younger Dryas event [29]. 3.2.2 Ice-rafted debris in the North Atlantic (32,000 yr BPÐpresent) Bond et al. have analysed sediments of ice rafted debris (IRD) in the North Atlantic [30, 31]. The latter are found in deep sea cores as layers of tiny stones and micro-fossils that were frozen into the bases of advancing glaciers and then rafted out to sea by glaciers. These reveal abrupt episodes when cool ice-bearing waters from the North Atlantic advanced as far south as the latitude of Britain, coincident with changes in the atmospheric circulation recorded in Greenland. A quasi-cyclic occurrence of IRD events has been found, with a periodicity of 1470 ± 530 yr, during which temperatures dropped and glacial calving suddenly increased (Fig. 6). The underlying cause of these events is not yet known but the evidence tightly constrains the possibilities. First, the rafting icebergs are launched simultaneously from more than one glacier, so the driving mechanism cannot be ascribed to a single ice sheet but requires a common climate forcing mechanism. It points to a trigger that caused air temperatures to drop and induce the release of ice over a large region. Second, the events continue with the same periodicity through at least three major climate transitions: the Younger Dryas-Holocene transition, the deglaciation, and the boundary within the ice age between the marine isotope stages 2 and 3 (Fig. 6) [31]. Even though the ice conditions during these transitions were changing dramatically, the IRD events continued with the same periodicity. Third, and especially surprising, is the evidence that the IRD cold events have continued through the Holocene (Fig. 6), with the same periodicity (but of course with a lower amount of IRD material). The events were abrupt during both the glacial and Holocene periods, generally switching on and off within one or two centuries. The estimated decreases in North Atlantic Ocean temperatures during the Holocene IRD events are 2 K, or about 15–20% of the full Holocene-to-glacial temperature difference. This observation questions the validity of the currently-held picture that the Holocene has been a period of exceptional climatic stability—and much more stable than previous interglacials. For the North At- lantic at least, the IRD data show that there has been much more climate change during the Holocene than previously thought. The implication of these observations is the presence of a quasi-periodic climate cycle of about 1500 yr that occurs independently of the glacial-interglacial climate state. Furthermore, the IRD peri- odicity is suggestive of the pacing of the warm Dansgaard-Oeschger events during the ice age. These events (which are seen in stage 3 of Fig. 6a) are abrupt warmings of Greenland by about 5–10 K over afew decades, followed by gradual cooling over several hundred or thousand years. Presumably the warming of the waters far north leads to an increased calving of glaciers. Simulations [32] suggest that the cold stadial periods are the ‘stable’ mode of the glacial Atlantic Ocean circulation, with NADW for- mation south of Iceland—the so-called ‘cold’ conveyor mode. The warm Dansgaard-Oeschger events represent a temporary transition to the ‘warm’ conveyor mode with NADW formation further north, in the Nordic Seas. A small decrease of freshwater into the North Atlantic is sufficient to trigger these events. What causes these changes in freshwater production is not yet know, although there is increasing evidence that solar forcing is involved. Until recently the origin of the quasi-1500 yr climate cycle was unknown. Ice sheet oscillations are ruled out as the forcing mechanism. Orbital periodicities around the Sun are too long to cause millennial- scale climate cycles. However a recent study has shown that solar variability is highly correlated with the ice rafted debris events during the Holocene (Fig. 7) [34]. The correlation embraces the Little Ice Age, which appears to be the most recent of these events. This rather convincing evidence implies that solar forcing has caused at least the Holocene section of the quasi-1500 yr climate cycle in the North Atlantic. It seems likely that solar forcing also caused changes in the hydrological cycle and North Atlantic Deep Water production, triggering the Dansgaard-Oeschger events and providing an additional mechanism for globally amplifying the solar signals. (Note that although the Dansgaard-Oeschger events show a strong correlation with decreased 10Be concentration in Fig. 6 , this is largely due to increased rainfall (dilution) rather than a change the production rate (GCR intensity) [33].) Holocene Stage 2 Stage 3 a) -34 13589264 7 10 11 12 1314 15 16 17 Interstadials -36 (warm events) -38 GISP2 ice core O (per mil) 18 -40 δ -42 Ice-rafted debris 6 YD (hematite-stained grains Little Ice Age to 18 Holocene event 1 GISP2 interstadial in δ O 4 b) c) 2 Event pacing (kyr) 0 0 20 40 60 Calendar age (kyr BP) Fig. 6: Timing of ice-rafted-debris events in the North Atlantic [31]. The curves are a) the GISP2 Greenland ice core δ18O record showing Greenland temperatures for the Holocene, the late glacial (Stage 2) and the mid glacial (Stage 3) periods, b) the periodicity of the ice rafted debris events from 32 kyr BP to the present, measured from haematite-stained grains (other tracers give similar results) and c) the periodicity of Dansgaard-Oeschger warm events in the GISP2 δ18O data from 58 kyr to 26 kyr BP. a) 0.2 ice-rafted debris 14C ) 4 -1 s -2 0 0.0 C production rate 14 (atoms cm -4 LIA r = 0.44 Smoothed Combined ice-rafted debris (%) -0.2 0 2 4 6 8 10 12 Calendar age (kyr BP) b) 0.8 ice-rafted debris 10Be 4 ) -2 Be flux 0 0.0 10 atoms cm 5 10 -4 x ( Smoothed r = 0.56 Combined ice-rafted debris (%) -0.8 0 2 4 6 8 10 12 Calendar age (kyr BP) Fig. 7: Correlation of solar variability with ice-rafted debris events in the North Atlantic during the Holocene [34]: a) the 14C record (correlation coefficient 0.44) and b) the 10Be record (0.56), together with the combined ice-rafted-debris tracers. The Little Ice Age (LIA) is labelled in the upper figure. 3.2.3 Lake levels in the Jura Mountains (12,000 yr BPÐpresent) Magny has reconstructed the history of lake-Levels in the French Jura Mountains which lie near the French-Swiss border [35]. These correlate well with the ∆14C (GCR) record over the last 10 kyr (Fig. 8). This of course implies that the Jura lake levels also correlate with the periods of increased ice rafted debris in the North Atlantic [29] (§3.2.2). Fig. 8: Lake-Levels in the French Jura Mountains, and the ∆14Cvariation over the last 12 kyr. (Fig. 8) [35]. a ∆14 10 C (GCR intensity) –3.5 5 –4.0 0 –4.5 C (‰) –5 14 ∆ O (‰ VPDB) 18 –10 –5.0 δ increasing GCR deccreasing rainfall –15 –5.5 δ18O (rainfall) 6.500 7.000 7.500 8.000 8.500 9.000 9.500 Age (kyr BP) 10 b ∆14C (GCR intensity) 5 –4.0 0 –5 –4.5 C (‰) 14 –10 ∆ O (‰ VPDB) –5.0 18 δ increasing GCR –15 deccreasing rainfall –20 δ18O (rainfall) –5.5 7.9008.000 8.100 8.200 8.300 Age (kyr BP) Fig. 9: Profiles of δ18O from a U-Th-dated stalagmite from a cave in Oman, together with ∆14C from tree rings, for a) the 3.4 kyr period from 9,600 to 6,200 yr BP and b) the 430 yr period from 8,330 to 7,900 yr BP [36]. 3.2.4 Oman rainfall (9,600Ð6,200 yr BP) Neff et al. [36] have recently measured the δ18Ocontent in the layers of a stalagmite from a cave in Oman, which are U-Th dated to cover the period from 9,600 to 6,200 yr BP. The δ18Oismeasured in calcium carbonate, which is expected to be deposited in isotopic equilibrium with water. The data are shown in Fig. 9 together with the ∆14C obtained elsewhere from tree rings. The two timescales have been tuned to match bumps within the known experimental errors (smooth shifts have been applied to the U- Th dates up to a maximum of 190 yr). During a 430-yr period centred around 8.1 kyr BP, the stalagmite grew at a rate of 0.55 mm/yr—an order of magnitude faster than at other times—which allowed a high resolution δ18O measurement to be made (Fig. 9b). It is interesting to note that this coincides with an ice rafting debris cold event in the North Atlantic (§3.2.2). Oman today has an arid climate and lies beyond of the most northerly excursion of the inter tropical convergence zone (ITCZ), which carries with it the heavy rainfall of the Indian Ocean monsoon system. However there is evidence that the northern migration of the ITCZ reached higher latitudes at earlier times and, in consequence, that Oman had wetter climate. In this region, the temperature shifts during the Holocene are estimated to account for only 0.25 per mil variation in δ18Oof[36]. However the δ18O values of monsoonal rainfall associated with the ITCZ show an inverse correlation with rainfall and so, for these data, the δ18Ovariations are ascribed to changes of rainfall, as indicated in Fig. 9. Notice that the sign of the correlation is different from the other examples presented here; in this case a high GCR intensity is associated with a low rainfall. However, it is entirely plausible that a climate change can lead to different responses in different regions of the Earth. For example a globally-averaged increase of rainfall may still result in a decrease in certain regions due to effects such as a shift of the ITCZ. The similarity between the δ18O and ∆14Ccurves in Fig. 9, both in their long-term and short-term variations, is striking. It suggests that solar-GCR activity controlled the pattern of tropical rainfall and monsoon intensity during this 3,000-year period on decadel to centennial timescales. 20 15 ) -3 10 5 C (x 10 14 ∆ 0 -5 1000 900 800 700 600 500 Year (BC) 50 sphagnum 0 imbricatum 50 0 sphagnum papillosum 20 sphagnum cuspidatum 0 50 sphagnum secf. acutifolia 0 20 corylus avellana (pollen percentages) rcentage by volume (%) volume rcentage by 10 Pe 0 -37 0 Relative depth (cm) Fig. 10: Transitions in the fauna of a Netherlands peat bog core for the period 1000–500 BC [37]. Also shown is ∆14C for this period. At the onset of the rise in 14C around 800 BC, the peat-forming mosses shifted from those preferring relatively warm conditions to those preferring colder and wetter conditions. 3.2.5 Netherlands peat bog fauna (1000Ð600 BC) Van Geel et al. [37] have studied peat-forming mosses in raised bogs in The Netherlands that were laid down in the period 1000–500 BC. They find an abrupt shift occurred around 800 BC from mosses preferring relatively warm conditions to those preferring colder and wetter conditions (Fig. 10). This coincides with a sharp rise in ∆14C due to a decrease in solar activity. There is supporting evidence of a substantial climate shift at that time from archaeological remains of nearby Bronze Age settlements which had been continuously inhabited for more than a thousand years but were abandoned around that time, presumably as the ground became waterlogged. There is extensive evidence that this solar-induced change to a colder and wetter climate around 800 BC was a global phenomenon. Some examples are as follows. Migrations of settlements are recorded in central Asia at this time [37]. An ice rafting debris event (§3.2.2) occurred in the North Atlantic around 2800 yr BP [31]. A substantial glacier advance took place at this time in the presently-arid south-central Andes Mountains of northern Chile, which has been attributed to a marked increase of precipitation [38]. Arecent study of stalagmite growth rates in caves in the south-western United States shows that the period from 2800 to 2600 yr BP was the wettest for this (presently semi-arid) region during the last 4000 yr [39]. 40 a) Wolf Spörer Maunder minimum minimum minimum 20 Suess C (‰) 0 14 ∆ -20 Medieval Warm Little Ice Age 30 b) 20 10 Lake depth (m) 0 1000 1200 1400 1600 1800 2000 Calendar year (AD) Fig. 11: a) History of ∆14C from tree-ring analyses for the last millennium [40]. Recorded periods of climate change are indicated. The sharp negative 14Cdeviation during the present century is the Suess effect, due to the burning of 14C-depleted fossil fuels. b) History of rainfall and drought in equatorial East Africa during the last 1100 years [43]. The figure shows the reconstructed depth of Crescent Island Crater Lake, Kenya. The radiocarbon dating error for the lake data is ±50 yr. 3.2.6 Kenyan lake levels (900Ð2000 AD) The ∆14Cdata for the last 1000 years reveal considerable solar variability (Fig. 11a) [40]. The periods of large 14Cdeviation correspond to recorded climatic anomalies: a) 1000–1270, the so-called Medieval Warm period, b) 1280–1350, the Wolf Minimum, c) 1420–1540, the Sporer¨ Minimum, and d) 1645– 1715, the Maunder Minimum. Temperatures during the Medieval Warm epoch were elevated above normal, causing severe and extended droughts for the Anasazi in the south-western United States but allowing the Vikings to colonise Greenland and wine-making to flourish in England. It was followed by a period of about 4 centuries during which—save for a few short interruptions—the glaciers advanced and a cooler, harsher climate predominated. During this so-called Little Ice Age the River Thames in London regularly froze across, and fairs on the ice were a standard winter feature. The Little Ice Age was recorded in many parts of the world. For example, in China the rice crops of the Yellow River Valley were reduced from two to only one a year. Stalagmite studies in the south-western United States [39], Madagascar [41] and Nepal [42] have also recorded the Little Ice Age. Evidence has also been found that the Medieval Warm and Little Ice Age climates extended into the equatorial regions, providing further support that they were global phenomenon. Figure 11b) shows the correlation of the 14C record with the depth of a lake in equatorial East Africa over the last 1100 years [43]. The reconstruction is based on three independent palaeolimnological proxies: sediment stratigraphy and species compositions of fossil diatoms and midges. These data not only confirm the presence of the major climatic anomalies associated with the Medieval Warm period and the Wolf, Sporer¨ and Maunder Minima but also identify three extended drought periods between the minima: AD 1390–1420, 1560– 1625 and 1800–1840. The cultural history of this region, preserved in records and oral tradition, has recorded alternating periods of drought and prosperity that coincide with the lake-level reconstruction. 3.2.7 Ionian Sea sediments (1750Ð1975 AD) Biological organisms frequently preferentially use light isotope species because of the lower internal energy ‘costs’ to the organism associated with breaking the bonds in these molecules—so-called ki- netic isotope fractionation. The result is significant fractionation between the substrate (heavier) and biologically-mediated product (lighter). The magnitude of the kinetic fractionation depends on reaction rates, concentrations of products and reactants, and environmental conditions such as light and tempera- ture. 200 0.15 δ13C in ocean sediments ) -3 0.10 sunspots 150 0.05 100 0.00 Sunspot number C component (x10 50 13 δ -0.05 -0.10 0 1750 1800 1850 1900 1950 2000 Year Fig. 12: The δ13C content of Globigerinoides ruber skeleton sediments in the Ionian Sea, Italy, over the period 1750–1975 [44] (thick line). Also shown is the sunspot series over the period 1750–1995 (thin line). The δ13C amplitude is obtained from the 11.3 yr component found in a singular spectrum analysis (SSA). One such study that has been made is the δ13Ccontent of skeleton sediments of Globigerinoides ruber,asymbiotic planktonic foraminifera, in the Ionian Sea, Italy [44]. The δ13Cvariations in symbiotic foraminifera mainly measure the symbiont density and the photosynthetic activity, which varies with incident light level. Analysis of the time series for the period 1147-1975 AD has revealed an 11-year component, with high significance, that is in phase with the solar cycle and has an average amplitude of 0.04 per mil. The data for the last 250 yr period are shown in Fig. 12 and show a higher amplitude of about 0.08 per mil in recent solar cycles. Estimates indicate that this amplitude is compatible with the variation of sunlight expected from a relative change in cloud cover of about 3% over the solar cycle [44]. This is consistent with the satellite-observed value of 1.7% / 61% = 2.8%. It is interesting to note that the solar modulation of Globigerinoides ruber continued through the Maunder Minimum when the sunspots disappeared (and with them, disappeared also the sunspot /facular irradiance modulation). However the solar cycle modulation of 10Be continued through the Maunder Minimum, as shown by the 10Be ice core measurements [45]. These observations are consistent with a GCR interpretation for the change of cloud cover, but not with a solar irradiance variation interpretation (§4.1). 3.3 Solar-GCR-climate change in the Industrial Age 3.3.1 Solar-GCR change The variation of 10Be concentration in the Greenland ice core over the last 300 years (Fig. 13) [46] reveals considerable changes of solar magnetic activity have occurred in recent times. The peaks in GCR 0 Solar open magnetic flux Wb) 14 1.6 5 1.4 1.2 10 atoms/g) 4 1.0 0.8 Solar magnetic flux (x 10 15 0.6 0.4 10Be concentration Be concentration (x 10 Be concentration 0.2 Maunder Dalton 10 minimum minimum 0 1700 1750 1800 1850 1900 1950 2000 Year Fig. 13: Variation of 10Be concentration in the Greenland ice core over the last 300 years [46], due to changes of solar magnetic activity (thin line). The variation of the solar coronal source flux, FS, over the last 140 years is shown by the thick line [11] (note inverted scale and unsupressed zero). In the period since 1901, the increase of the solar open magnetic flux has been a factor 2.3. intensity over this period coincide with the cold spells of the Maunder Minimum and Dalton Minimum. More recently, during the 20th century, the 10Be data show a substantial reduction of the GCR intensity due to a marked increase in the strength of the solar wind. The latter is independently confirmed by the 6The 25 low cloud fraction 1.5 20 observed low cloud fraction 15 1.0 10 0.5 5 GCR intensity 0 0 GCR intensity variation (%) GCR intensity variation -5 -0.5 -10 (%) variation cloud fraction Global low -15 -1.0 1860 1880 1900 1920 1940 1960 1980 2000 Year Fig. 14: Estimates of the variations of GCR intensity (3 GeV/c cutoff) and low cloud absolute fraction from 1870 to 2000 (thin line) [16]. The observed variation of low cloud fraction for the period 1983-1994 is also shown (thick dashed line). The estimates are based on the coronal source flux (FS) measurements over this period obtained from the There is some evidence that solar forcing may also trigger oscillations in the Earth’s climate sys- tem, perhaps as secondary processes after changes in the clouds and hydrological cycle. Evidence has been discussed above for a solar forcing of the thermohaline circulation (§3.2.2). Another example may be the anomalous warming (El Nino)˜ or cooling (La Nina)˜ of the surface water in the eastern equatorial Pacific Ocean, which occurs in conjunction with the Southern Oscillation, a see-sawing of atmospheric pressure between the eastern and western tropical Pacific. The combined El Nino˜ Southern Oscillation (ENSO) together with La Nina˜ is the strongest source of natural variability in the Earth’s climate system on short timescales, and dominates short-term global temperature anomalies. ENSO is widely viewed as an example of a free internal oscillation of the Earth’s climate system, independent of any external forcing. However, Landscheidt [48] has presented evidence that the timings of El Nino,˜ La Nina˜ and the Southern Oscillation over the last 50 years can be linked with phases within the ascending and descending parts of the solar cycle. 3.3.2 Climate sensitivity to radiative forcing Climate model calculations indicate an approximately linear relationship between global mean radiative forcing, ∆F (Wm−2), and the equilibrium global-mean surface temperature change, ∆T (K), ∆T = λ ∆F (2) where λ (K/Wm−2)isthe climate sensitivity parameter. This parameter is relatively insensitive to the nature of the forcing, for example, greenhouse gases or solar irradiance, provided the forcing agent is not highly variable spatially (like, for example, aerosols). All climate feedback processes, such as changes in water vapour, clouds or ice sheet albedo, are implicitly included in λ. The value of λ can be inferred from past climate change and from climate models. For example, using ice core samples, between glacial and interglacial periods it is estimated that λ 5K/7Wm−2 −2 = 0.7 K/Wm .Climate models indicate a doubling of the concentration of atmospheric CO2 from pre- industrial levels (280 ppm) produces +4 Wm−2 forcing and a mean temperature rise ranging from 1.5 K to 4.5 K, with a central value of 3 K [49]. Therefore λ (3 ± 1.5)/4=(0.75 ± 0.4) K/Wm−2,in agreement with the previous estimate. These figures can be compared with the response of the Earth if it were to act as a simple black body. In this case the radiant emittance is R = σT 4, where σ is the Stefan-Boltzmann constant. The radi- ation from a black body varies as ∆R/R =4∆T/T,sothat ∆T =(T/4R)∆R. Since ∆R/R =∆I/I, the fractional change in solar irradiance, it follows that λ0 = T/4I. The effective radiating temperature of the Earth is T 266 K, and the global mean solar irradiance reaching the lower troposphere is I 0.7 × 1366/4 240 Wm−2 (the factor 0.7 accounts for shortwave albedo and the factor 4 aver- ages the solar irradiance of 1366 Wm−2 over the full surface area of the Earth). We thereby estimate −2 λ0 = 266/(4 · 240) 0.3 K/Wm for the Earth in the absence of any feedbacks. Therefore the climate feedback factor of the Earth is between a factor of about 1.2 and 4, with a central value of 2.5, and it is greater than one, amplifying the temperature response to a radiative forcing compared with that for a simple black body. 3.3.3 Solar contribution to the current global warming The reconstructed global mean surface temperature of the Earth between 1860 and 2001 (Fig. 15) [49] indicates a warming of about 0.6 K over this period. A notable feature of the warming is that it did not rise smoothly along with the steadily increasing emissions of anthropogenic greenhouse gases but seemed to flatten, or even reverse sign, during the period 1945–1980. 0.8 GLOBAL AVERAGE C) 0 0.4 0.0 0.4 Temperature difference ( relative to 1961-1990 average Data from thermometers 0.8 1860 1880 1900 1920 1940 1960 1980 2000 Year Fig. 15: The global mean surface temperature of the Earth, 1860–2001, relative to the 1961–1990 average [49]. A component of this temperature reconstruction is sea-surface temperatures (SSTs), which have been measured on a routine basis by ocean-going ships since the mid-19th century. The SST record is a particularly valuable measure of global climate since it represents over 70% of the Earth’s surface and is much more spatially and temporally homogeneous than the land surface, as well as being free of such problems as the warming from ‘urban heat islands’. The mean SSTs over the period 1860–1985, for the Atlantic, Pacific and Indian Oceans, are shown in Fig. 16, together with the global mean SST [50]. All of these oceans show a temperature rise that levels off in the same period around 1945–1980, as well as a cooling around the beginning of the last century. Both of these features are characteristic of solar activity, as can be seen in the smoothed sunspot number (Fig. 16) and in the GCR intensity (Fig. 17). Aworld-wide simultaneous variation of SST puts severe constraints on a possible forcing mech- anism. Since the same characteristic features are seen in all oceans, they are unlikely to be caused by changes such as El Nino˜ events, shifts in wind patterns or changes in the thermohaline circulation, which would lead to differences between the oceans. The mechanism could in principle be increases of anthro- pogenic greenhouse gases, but the variation in the first half of the 20th century occurred before these were significant. There were insufficient volcanic events to account for the mid-century cooling. The inescapable conclusion is that these data provide quite strong evidence that solar variability was the pri- mary cause of the warming during at least the first half of the last century. It remains an open question as to what was the solar contribution to the warming during the second half of the 20th century. 90 0.15 80 0 -0.15 70 Atlantic Ocean Sunspot number -0.30 60 -0.45 0.2 50 0 40 -0.15 0.1 Pacific Ocean Sunspot number (smoothed) -0.30 0 30 0.30 Temperature (K) -0.1 0.15 -0.2 0 Temperature (K) -0.15 -0.3 Global mean SST Indian Ocean Temperature (K)-0.30 Temperature (K) -0.4 1860 1880 1900 1920 1940 1960 1980 1860 1880 1900 1920 1940 1960 1980 Year Year Fig. 16: Annual mean sea-surface temperatures (SST), 1860–1985, for the Atlantic, Pacific and Indian Oceans (right-hand panel) and the global mean (lower curve in the left-hand panel) [50]. The temperatures are shown relative to their 1951–1980 averages. Also shown is the 11-year running mean of the annual sunspot numbers (upper curve in the left-hand panel). The smooth curves are 7th order polynomial fits to the data. 0.3 -10 0.2 -5 0.1 0 0 temperature anomaly -0.1 5 -0.2 cosmic ray intensity -0.3 10 emperature anomaly (oC) emperature (> 3 GeV/c) T (11-y running mean) GCR intensity change (%) -0.4 15 1860 1880 1900 1920 1940 1960 1980 2000 Ye ar Fig. 17: The mean GCR intensity (3 GeV/c cutoff) over the period 1870–1990, smoothed with an 11-year running mean, together with the global mean temperature anomaly in the same period. The GCR intensity is based on the correlation between directly measured values and the coronal source flux estimates of Lockwood et al. (§3.3) [11]. 3.3.4 Solar signal in the temperature record If there is a solar contribution to the current global warming then a solar cycle signal should be present in the temperature record. However the temperature variation is expected to be quite small. The 1.1 Wm−2 (0.08%) variation of the solar irradiance (§4.2) corresponds to a global mean variation of 0.7 × 1.1/4= 0.2 Wm−2 at the Earth’s surface. This would be expected to produce an equilibrium temperature change ∆T =0.7 × 0.2=0.14 K (i.e. a sinusoidal amplitude of 0.07 K). However the large thermal mass of the surface layer of the oceans will reduce the actual temperature response. Table 2: Estimation of the ocean’s temperature response to a solar cycle (11-yr sinusoidal) forcing of amplitude 0.1 Wm−2, using an RC-circuit equivalent (§3.3.4). The attenuation factor is the ratio of the maximum amplitudes for an 11-yr sinusoidal forcing to a constant forcing. Climate Time ∆Tfor constant Attenuation Solar T Phase sensitivity, λ constant forcing factor amplitude lag (K/Wm−2) (yr) (K) (K) (degrees) 0.3 3.6 0.03 0.43 0.013 64 0.5 6.0 0.05 0.28 0.014 74 0.7 8.4 0.07 0.20 0.014 78 Fig. 18: Sea surface temperature anomalies (Kelvin) from bathythermograph measurements collected from 1955 to 1994 [52]. The light curves in the upper four panels show monthly mean values and the heavy curves show low-pass filtered values (with half-power points at 7 yr). The lowest panel shows the reconstructed solar irradiance over the same period, including the low-pass filtered values (the left hand axis gives the value at the top of the atmosphere and the right hand axis gives the global mean values at the sea surface). We can estimate the damping effect of the oceans by considering an analogous electromagnetic RC-circuit equivalent for the ocean surface layer: a resistor and capacitor in series, which are set into forced oscillation by a sinusoidal voltage [51]. In this analogy, forcing heat fluxes are analogous to current, temperature to voltage, the ocean to a capacitor and the climate sensitivity parameter to a resistor. The value of the capacitor can be estimated by assuming the ocean can be represented by a well-mixed upper layer of about 90 m depth that is effectively isolated from thermal exchange with deeper water, except by relatively slow diffusion. Table 2 summarises the expected ocean response to a solar cycle (11-yr sinusoidal) forcing of amplitude 0.1 Wm−2. The estimated solar cycle temperature amplitude is 0.014 K, with a phase lag of about 65–80 degrees. The temperature response is essentially independent of the climate sensitivity, λ,since the larger constant-forcing temperatures at higher λ are compensated by longer time constants and therefore larger attenuation factors for the short solar cycle. The attenuation factors are large, e.g. a factor 5 for λ = 0.7 K/Wm−2, corresponding to an 8.4 yr time constant. White et al. [52, 53] have analysed the SST data for the period 1900–1991 and the bathyther- mograph data for 1955–1996. Their analyses reveal convincing solar signals in the Indian, Pacific and Atlantic Oceans which all show comparable amplitudes and phases (Fig. 18). The solar cycle amplitude is (0.03±0.005) K and it lags the changes in solar irradiance by about 0–65◦. This amplitude is twice the expected value (Table 2). An inter-decadel (18–25 yr) solar signal is also observed with (0.04±0.005) K amplitude and 15–50◦ phase lag, which is consistent with the expected value from estimated longer-term changes in solar irradiance [53]. In summary, there is indeed clear evidence of a solar signal in the tem- perature record but the 11-yr oscillation appears to be about a factor two larger than expected, suggesting either an error in the modelling, or else the presence of a mechanism that amplifies the solar forcing. anthropogenic natural natural greenhouse gasesland use aerosols & clouds Sun & volcanoes GCR-cloud? +2 a) b) total +2.9 W/m2 -0.2 -1.4 (-2.4) W/m2 +0.4 W/m2 +0.8 W/m2 (solar) ? ) 2 +1 land aerosol aerosol indirect trop. albedo direct cloud cloud albedo lifetime 0 strat. ? solar solar direct indirect CO2 CH4 CFC O3 N20 volcanic Radiative forcing (W/m -1 aerosol (range of 10y mean) -2 Fig. 19: a) The global mean radiative forcings of the climate for the period from pre-industrial (1750) to present, as estimated by the Intergovernmental Panel on Climate Change (IPCC) [49]. A positive forcing causes a global mean warming, and a negative forcing causes a cooling. The aerosol indirect contributions are poorly known, especially the second one (changes of cloud lifetime), and have large uncertainties. b) Estimated radiative forcing due to the putative solar indirect effect if GCRs and low cloud cover are causally linked. 3.3.5 Radiative forcings The global mean radiative forcings of the climate in the Industrial Age, as estimated by the Intergov- ernmental Panel on Climate Change (IPCC), are shown in Fig. 19a) [49]. Whereas the forcings due greenhouse gases are known quite well, there are large uncertainties associated with the forcings due to anthropogenic aerosols. The latter are separated into aerosol direct effects (albedo and absorption of solar radiation) and aerosol indirect effects due to their influence on clouds (the so-called first indirect type is the change in cloud albedo and the second indirect type is the change in cloud lifetime). All three of these processes are estimated to contribute a net negative forcing. The change in cloud lifetime is indicated in the figure as being of roughly equal importance as the change in cloud albedo. However both of the indirect aerosol effects are poorly known, and, moreover, new effects have recently been reported [54] that make their large uncertainties even larger. These relate to the effects of surface tension depression of water droplets by anthropogenic organic surfactants, which lead to an increase of droplet number concentration at a given water vapour supersaturation (§4.4). Ignoring cloud lifetime changes, the total estimated anthropogenic forcing is about +1.3 Wm−2. In comparison, the estimated solar direct contribution is +0.4 Wm−2. These figures can be converted to − an approximate expected change in equilibrium temperature using Eq. 2: ∆T (K) = 0.7 · ∆F (Wm 2). However Fig. 19 takes no account of the spatial and temporal distribution of the forcings, which are highly non-uniform in the case of aerosols, and so the actual temperature response may be quite different. Nevertheless it is probably reasonable to conclude from Fig. 19a) that the residual anthropogenic forcing is the small difference between two relatively large numbers—a positive forcing from greenhouse gases and a negative forcing of uncertain magnitude from anthropogenic aerosols. If GCRs are indeed the causal mechanism for the observed changes in cloud cover, then we can estimate the resultant forcing since 1900 to be about +0.8 Wm−2 (§3.3.1), as shown in Fig. 19b). This so- lar indirect effect is potentially a sizeable forcing—about a factor two larger than the supposed changes in solar irradiance over the same period, and with the same sign. Together with the previous contribu- tions, this would imply a total mean forcing during the 20th century of about +1.3 Wm−2 anthropogenic and +1.2 Wm−2 natural (solar). From these figures should be subtracted the cooling effects of the an- thropogenic increases in cloud lifetimes, and of volcanoes, respectively. If we take the central value for the climate sensitivity, λ = 0.7 K/Wm−2, then it would imply a larger warming than the 0.6 K observed. This discrepancy could be due to several reasons. On the one hand some of these contributions are poorly known and their estimated magnitudes may change. This includes the possibilities that there is a signif- icant anthropogenic effect on cloud lifetimes and, of course, that the GCR-cloud effect may not exist. Alternatively, or in addition, the climate sensitivity parameter may be less than 0.7 K/Wm−2,inwhich case the projected anthropogenic temperature increase during the present century would be reduced. In conclusion significant uncertainties remain in estimating the radiative forcings from anthro- pogenic and natural sources. The largest uncertainties concern the microphysics of clouds and aerosols. Among these is the possible new contribution due to cosmic ray-cloud interactions. It is clearly important to either confirm or rule out this hypothesis as a natural mechanism for climate change. In the remain- der of this paper we will first look at the possible microphysical mechanisms that could be responsible for solar-cloud variability and then describe the proposed CLOUD experiment to test the GCR-cloud mechanism. 4 PHYSICAL MECHANISMS FOR SOLAR-CLOUD VARIABILITY 4.1 The Sun-Earth link There are only three physical paths that could connect variations of the Sun to the Earth’s climate (since neutrinos can be safely ignored!): 1. Solar electromagnetic radiation. 2. Galactic cosmic rays, whose intensity is modulated by the solar wind. 3. Solar wind, and its direct interaction with the troposphere. The third option is probably not important since the charged particles of the solar wind generally have very low energy (few keV) and so they are easily shielded by the Earth’s magnetosphere—that is except over the polar regions, where they range out in the thermosphere at an altitude of about 100 km. This is far from the tropopause (which lies at an altitude of about 8 km over the poles and 18 km over the tropics). Large coronal mass ejections (CMEs) that aim towards the Earth’s magnetosphere can generate severe magnetic disturbances and cause electron precipitation events in the polar regions. There is an appreciable rate of such events, about 60–80 per year, concentrated about 3 years after the peak of the solar cycle, i.e. with approximately the opposite solar phase as that of the GCR flux [55]. These few- MeV electrons reach altitudes of about 25 km over the polar regions and can influence processes in the the polar stratosphere. Occasionally, very energetic CMEs give rise to so-called solar cosmic rays (SCRs; also known as solar energetic particles, SEPs) of a few ×100 MeV maximum energy, which are thought to be generated by a linear shock acceleration mechanism. These may penetrate to ground level at high geomagnetic latitudes. They are relatively rare, however, occurring at a peak rate of about 3–8 per year around solar maximum (preferentially during the rising and falling part of the cycle), with almost none around solar minimum [55]. During these infrequent events, SCRs could affect the atmosphere via the same microphysical interactions as GCRs. In summary, there are only two plausible paths that could connect variations of the Sun with the Earth’s global clouds, namely: 1) solar electromagnetic radiation and 2) GCRs, via solar-wind modula- tion. We will consider these two candidates in more detail below. Days (since 1 Jan 80) 0 2000 4000 6000 1369 IM II HF HF HF CRIM I CRIM I VIRGO A A 1368 ACR ) 2 1367 0.1% 1366 1365 Solar irradiance (Wm Solar irradiance 1364 1363 78 80 82 84 86 88 90 92 94 96 98 Year Fig. 20: Total solar irradiance at the top of the Earth’s atmosphere over the last two solar cycles [9]. Sunspot maximum corresponds to peak irradiance. The relatively large and rapid fluctuations are due to sunspots rotating into and out of the field of view. 4.2 Solar electromagnetic radiation It is natural first to consider variations of the solar irradiance—either in overall intensity or in the dis- tribution of insolation in space and time—as a possible cause of climate change. Indeed there is strong evidence that the Milankovitch theory of climate forcing, due to variations of the Earth’s orbit around the Sun, plays an important role in long-timescale (10–100 kyr) climate change. Milankovitch identified three types of orbital variation that could act as climate forcing mechanisms: tilt of the Earth’s axis, precession of the equinoxes, and eccentricity of the Earth’s orbit around the Sun. Each has its own char- acteristic periodicity and phase, and these are seen in palaeoclimatic studies. Nevertheless, it seems that orbital forcing mechanisms alone could not account for the magnitude of the observed climatic variations over the past 2 million years. Other mechanisms—such as positive feedbacks or perhaps entirely new mechanisms—need to be invoked An additional possibility is a variation of the solar irradiance itself. Satellite data (Fig. 20) [9] have shown that the total solar irradiance is indeed varying over the course of solar cycle, but by a tiny amount of about 0.08%. This can be quantitatively well-explained by sunspot darkening and facular brightening O2 O3 H2O & CO2 UV VIS IR ) 4 ~30km altitude cutoff -1 10 spectral irradiance nm -2 (top of atmosphere) 102 0 km 100 1.0 0.1 -2 10 spectral 0.01 variability 0.1% 0.001 (max-min)/max Solar irradiance (mW m Solar irradiance 10-4 Solar cycle variation 102 103 104 105 Wavelength (nm) Fig. 21: Wavelength dependence of the solar irradiance at the top of the atmosphere, and its variation from sunspot minimum to maximum. Also shown is the solar spectrum at the Earth’s surface after absorption by the atmosphere. [56]. Together with measurements of cycling stars similar to our Sun, this had led to estimates of longer- term changes of solar irradiance that appear to be too small to account for the observed climate changes (§1). Attention has therefore focussed on changes of the ultra violet (UV) component of the solar spec- trum [57] which, although it carries only a small fraction of the total energy (about 0.1%), shows a much larger variation of several per cent over the solar cycle (Fig. 21). The UV wavelengths are absorbed at altitudes above 30 km by oxygen (<240 nm wavelength) and ozone (200–300 nm), and cause measurable heating of the thin atmosphere in the upper stratosphere. A positive feedback mechanism exists since the increased UV creates more ozone, although the fractional change is small (about 1–2% from solar minimum to maximum). Modelling reproduces these changes, and studies (e.g. ref. [58]) suggest that circulation changes initially introduced in the stratosphere by this heating can affect circulation at lower altitudes in the troposphere, and therefore can in principle influence cloudiness. It is clearly important to investigate this mechanism further with more experimental and modelling studies. 4.3 Cosmic rays, via solar wind modulation The other candidate link between solar variability and the Earth’s climate is via GCRs, which are mod- ulated by the solar wind. In contrast with solar UV radiation, GCRs directly penetrate the lower tropo- sphere where the cloud variation is observed, and they have an appreciable intensity variation over the solar cycle. 4.3.1 Solar wind characteristics: The solar wind is a continuous outward flow of plasma (mainly protons and electrons, with about 5% heavier ions) from the Sun’s corona. As a consequence of its high electrical conductivity, a weak magnetic field is ‘frozen’ into the plasma. The solar wind follows the Parker spiral trajectories out over the huge volume of the heliosphere to distances of 50–100 AU, well beyond the orbit of Nep- tune. At the Earth’s orbit it has a velocity of 350–800 km s−1 (β = 0.001–0.003), an intensity of (0.5–5)·108 particles cm−2 s−1, and a magnetic field of about 5 · 10−5 Gauss. The main sources of the solar wind on the Sun’s surface include large regions of open magnetic flux known as coronal holes; re- gions on the photosphere at the boundaries of the supergranulation cells, where magnetic reconnections occur; and coronal mass ejections. The sunspots themselves are but a visible indication of a high state of magnetic activity of the Sun, when the solar wind is strong. Sunspots are areas of the Sun’s photosphere where strong local a) b) poloidal sunspot c) d) flux tubes toroidal Fig. 22: Babcock’s model for generation of sunspot magnetic fields [59]. An initial weak dipole field (a) in the convective zone is wound up into a toroidal field (b) by the Sun’s differential rotation. Eventually the field becomes strongly toroidal (c) and magnetic flux tubes rise through the convective zone where they break through the photosphere to form sunspots with opposite polarity in the N and S hemispheres (d). 40 d solar rotation rate (nHz) 33 d 29 d rotation period convective zone Radius radiative zone tachocline core 25 d Radius Fig. 23: The interior rotation of the Sun from helioseismology measurements [60, 61]. The fastest rotation is 25 days at the equator and the slowest is over 35 days near the poles. Each contour is separated by 10 nHz (about 0.7 d). The tachocline is the shear region between the radiative and convective zones and is thought to be where the surface magnetic fields originate. magnetic fields emerge vertically. The fields are typically about 2500 Gauss, to be compared with a mean quiescent photospheric field of below a few Gauss. They appear dark because their temperature is about half of the surrounding photosphere (3,000 K compared with 5,800 K). They are generated (Fig. 22) by the differential rotation of the Sun with respect to latitude and depth: at the surface, one revolution takes 25 days at the equator and over 35 days near the poles (Fig. 23). This transforms the quiescent dipole field into a toroidal field and eventually creates ‘knots’ of strong localised fields. These knots may penetrate the photosphere to form sunspots, which appear cooler due to modification of the normal convective motions of the plasma by the strong magnetic fields. The sunspots first appear at high latitudes and then gradually migrate towards the equator. They eventually disappear by magnetic recombination, leaving a quiescent dipole field once more (but of opposite polarity). The half-cycle from dipole to toroidal and back to (reversed) dipole field is termed the solar, or sunspot, cycle and takes about 11 years on average. The key to solar variability is a fundamental understanding of the complex solar magnetic fields. How are they generated by the dynamo and what causes their quasi-periodic behaviour? Dynamo action involves the conversion of kinetic energy into magnetic energy by the inductive effects of fluid motion in an electrically conducting fluid—the solar plasma. Babcock’s qualitative picture of the solar dynamo (outlined in Fig. 22) has been known for over 40 years. But it has only been recently with the exquisite helioseismology measurements of GONG and other experiments, and with data from the high-precision spectrometers and detectors on board SOHO, Ulysses, Yohkoh, TRACE and other satellites, that great advances are being made. For example, it appears that the tachocline (Fig. 23) plays an essential role in the solar dynamo, and is the primary region for generation of the magnetic flux before it rises through the convective zone. Also, from high-resolution movies taken of the photosphere and corona, it appears that the dynamics of magnetic flux bundles, and their reconnections, are key to understanding the energy source that heats the solar corona and accelerates the solar wind. 4.3.2 Modulation of galactic cosmic rays by the solar wind: Cosmic rays are generated by supernovae and other energetic sources in our galaxy and beyond. On entering the heliosphere, charged cosmic rays are deflected by the magnetic fields of the solar wind. The transport problem of the GCRs through the heliosphere was first solved by Parker [62] and involves several processes of which the dominant is scattering off the magnetic irregularities, which produces arandom walk or diffusion effect. It has been shown theoretically [63] that the effect on the energy of a charged cosmic ray particle in passing through the heliosphere is equivalent to that produced by a heliocentric retarding electric potential with a magnitude at the Earth’s orbit equal to the energy lost by the cosmic rays in interacting with the solar wind. This retarding potential varies between about 1000 MV during periods of very high solar activity and zero during grand minima such as the Maunder Minimum. The solar wind therefore partly shields the Earth from the lower energy GCRs and affects the flux at energies below about 10 GeV. The effective retarding potential over the present eleven-year solar cycle averages about 550 MV, ranging from about 450 MV at the minimum to 850 MV at maximum. This leads to a distinct solar modulation of the GCR intensity (Fig. 24). The geomagnetic field also partially shields the Earth from GCRs. The dipole field imposes a minimum vertical momentum of about 13 GeV/c at the equator, 3 GeV/c at mid latitudes, and falling essentially to zero at the geomagnetic poles. In consequence, the GCR intensity is about a factor 3.6 higher at the poles than at the equator, and there is a more marked solar cycle variation at higher latitudes. Over the solar cycle, the variation of GCR intensity at the top of the atmosphere is about 15%, globally averaged, and ranges from ∼5% near the geomagnetic equator to ∼50% at the poles. At lower altitudes both the GCR intensity and its fractional solar modulation decrease. These are consequences of the absorption of low energy GCRs and their secondary particles by the atmospheric material, which totals about 11 nuclear interaction lengths. Balloon measurements (Fig. 25) show solar cycle variations of about 10% at low altitudes around 3 km (for a 2.4 GeV/c rigidity cutoff). solar maxima: cycle 19 20 21 22 Fig. 24: Balloon measurements of the cosmic ray intensity at shower maximum (15–20 km altitude) for the period 1957–1998, measured by the Lebedev Physical Institute. The curves correspond to four different locations for the balloon flights: Mirny-Antarctica (0.03 GeV/c rigidity cutoff), Murmansk (0.6 GeV/c), Moscow (2.4 GeV/c) and Alma-Ata (6.7 GeV/c). Due to atmospheric absorption, the data of Murmansk and Mirny practically coincide with each other. The approximate times of the sunspot maxima for the last 4 solar cycles are indicated. 0.095 0.70 solar maxima: cycle 19 20 21 22 8 km altititude ) ) 0.090 0.65 -1 -1 s s -2 -2 0.085 0.60 0.080 0.55 10% (3km) 0.075 3 km 0.50 GCR intensity at 3km, I (cm altititude Moscow GCR intensity at 8 km, I (cm (2.4 GeV/c rigidity cutoff) 0.070 0.45 1960 1970 1980 1990 2000 Year Fig. 25: Balloon measurements of the GCR intensity at 3 km and 8 km altitudes from 1957 to 2000 (2.4 GeV/c rigidity cutoff), measured by the Lebedev Physical Institute. The approximate dates of the sunspot maxima for the last 4 solar cycles are indicated. Altitude (km) 20 15 10 5 3 2 1 0 1 ) 10-1 -1 sr νµ + νµ -1 10-2 µ+ + µ− s -2 10-3 p + n 10-4 + − tical flux (cm e + e r -5 + − Ve π + π 10 Nuclear interaction lengths, λ 246810 10-6 0 200 400 600 800 1000 Atmospheric depth (g cm-2) Fig. 26: Vertical fluxes of cosmic rays and secondary particles (E>1 GeV) vs. altitude [64]. The primary − nucleons (p and n) include protons, He and heavier nuclei. The points show measurements of µ with Eµ >1 GeV. 4.3.3 Interactions of cosmic rays in the atmosphere: The composition of the charged primary cosmic rays at the top of the atmosphere is about 98% protons and heavier nuclei, and 2% electrons. Of the former, about 87% are protons, 12% are He nuclei and the remaining 1% are heavier nuclei (especially C, N, O and Fe). The incident nucleons interact with nuclei in the atmosphere and produce secondary particles in hadronic cascades. The initial secondaries are mainly p, n, π± and γ (from π0 decay), and these subsequently produce µ and e (Fig. 26). Below about 6 km altitude, muons from the decay of π mesons dominate the cosmic ray flux. The maximum cosmic ray fluxes occur at altitudes of 15–20 km, where the charged particle inten- sities vary between about 0.8 and 2.3 cm−2s−1 (at solar maximum, i.e. GCR minimum), depending on geomagnetic latitude (Fig. 27a) [65]. Most of the primary cosmic rays interact in the first ∼2λ material above the tropopause, so the heavily ionising primaries (heavy nuclei) are screened from reaching lower altitudes. Therefore, essentially throughout the troposphere, charged cosmic rays are mostly moderately relativistic singly-charged particles. These particles (other than electrons) lose energy primarily by ioni- sation of the air molecules, with an ionisation energy loss rate, dE/dx ∼ 1.7 MeV /g cm−2, characteristic of so-called minimum ionising particles.Inair at stp, the number of ion pairs produced by a minimum ionising particle is 60 ion-pairs cm−1 or, equivalently, the ionisation energy loss is 34 eV per ion pair created. At high altitudes near the GCR maximum the fraction of heavily-ionising non-relativistic parti- cles becomes significant and the mean ionisation density is about 110 ion pairs cm−1,corrected to one atmosphere pressure [65]. At 15 km the density of air is 0.20 × 10−3 gm cm−3 and the mean ionisation density is therefore about 18 ion-pairs cm−1 per charged particle. Therefore the ion pair production rate by cosmic rays at 15 km altitude is I = 18×(0.8–2.3) = (14–41) cm−3s−1, depending on geomagnetic latitude. At 3 km altitude, the GCR flux is about 0.08 cm−2s−1 at 2.4 GeV/c cutoff (Fig. 25), producing about 3.5 ion-pairs cm−3s−1.Atground level these are 0.02 cm−2s−1 and 1.2 i.p. cm−3s−1, respectively. Natural radioactivity also contributes to atmospheric ionisation over land. The relative contribu- tion from radioactivity and GCRs as a function of altitude is shown in Fig. 28. During rainfall, radon and its daughter radioisotopes are sedimented from the air, and the ionisation rate measured by plastic scin- tillation counters close to the ground can increase by up to about 25% [66]. This increase reflects only the γ ray component since α particles are generally not detected because of their short range. However, α particles represent an important component of the radon decay chain. Over oceans, the contribution of radioisotopes is negligible and so, averaged over the entire troposphere, GCRs are by far the dominant source of ionisation (more than 99%). 35 17.3 GeV/c 5.2 GeV/c 3.3 GeV/c 0.03 GeV/c 30 a) 25 20 15 10 5 0 0 0.5 1.0 1.5 2.0 2.5 Charged particle intensity (cm-2s-1) 35 17.3 GeV/c 5.3 GeV/c 3.3 GeV/c 0.03 GeV/c 30 b) 25 20 15 Altitude (km) Altitude (km) 10 5 0 01234 Negative small ion concentration (x1000 cm-3) Fig. 27: a) The charged particle intensity and b) the negative small ion concentration vs. altitude, measured at several latitudes with cutoff rigidities, Rc,asindicated. The data were recorded by Lebedev Physical Institute [65] in or near 1990, corresponding to a sunspot maximum (but without solar proton events), i.e. during a cosmic ray minimum. The horizontal bars show the typical experimental statistical errors. 40 radioactivity/GCR 30 relative ionisation (over land) 20 Relative fraction (%) 10 0 0123 Altitude (km) Fig. 28: Relative fraction of atmospheric ionisation from radioactivity and from GCRs as a function of altitude, over land. Free radicals are also created by galactic cosmic rays, which may lead to a significant source of chemically-reactive molecules in certain regions of the atmosphere. As examples, about 1–2 OH radicals [67] and 1.5 NO molecules [68, 69, 70] are estimated to be produced per ion-pair. Mixing ratios of about 1 pptv OH or NO are therefore generated by cosmic rays per day in the upper troposphere. 4.3.4 Evolution of ions in the atmosphere: The ions and free electrons created by cosmic rays rapidly interact with molecules in the atmosphere and + + convert to complex positive and negative cluster ions [71]. Primary positive ions are mostly N2 ,O2 , + + ∼ − N , and O . Free electrons rapidly (τ 200 ns) attach to O2,leading to O2 as the most important primary negative ion. Both positive and negative primary ions experience rapid ion-molecule reactions + with relatively abundant atmospheric gases, leading to the cluster ions H3O (H2O)n (formation time ∼ − ∼ 1msandn peaking around 4–6) and CO3 (H2O)n (formation time 10 ms). More complex cluster ions form on slightly longer timescales. The positive ions react further with + basic molecules B possessing proton affinities larger than that of H2O, leading to H B(H2O)n. Impor- + tant examples for B are ammonia, forming NH4 (NH3)m(H2O)n, and acetone (CH3)2CO. Negative ions − react with acidic molecules, particularly H2SO4 and HNO3, leading to HSO4 (H2SO4)l(HNO3)m(H2O)n − and NO3 (HNO3)m(H2O)n.The above species have been observed in the upper troposphere and lower stratosphere by aircraft-based ion mass spectrometers [72, 73]. As a consequence of the chemical dif- ferences between positive and negative ion clusters in the atmosphere, the latter have a slightly higher (20%) electrical mobility. These ion clusters form on timescales of order 1–100 s, depending on the trace gas concentrations. The formation time can be simply estimated as follows. At stp, the mean free path of air molecules is 0.067 µmand the rms velocity is 500 ms−1,sothe collision rate per molecule is about 500/(0.067 · 10−6) ∼ 1010 s−1. Therefore the mean time interval for a trace gas molecule to collide with an ion is about 0.3 s at 1 ppbv concentration (3 · 1010 molecules cm−3 at stp), and about 300 s at 1 pptv (3 · 107 molecules cm−3). The evolution of an embryonic cluster ion competes with ion “loss” mechanisms such as ion- ion recombination, ion-aerosol attachment and, in clouds, ion-droplet attachment. Away from clouds, the production rate of ions by cosmic rays, I [ion-pairs cm−3s−1]isinequilibrium with the loss rate according to I = αn2 + βnN (3) −3 3 −1 where n [cm ]isthe small ion concentration of one sign (and we assume n n+ n−), α [cm s ] is the ion-ion recombination coefficient (about 1.6 × 10−6 cm3s−1), β [cm3s−1]isthe ion-aerosol at- tachment coefficient (which varies with aerosol size and charge) and N [cm−3]isthe aerosol number concentration. If we assume for the moment that the principal removal mechanism is ion-ion recom- bination, then the expected equilibrium ion density (of one sign) at 15 km altitude is n = I/α = (14 − 41)/(1.6 × 10−6)=(3− 5) · 103 cm−3. The measured negative small ion concentrations vary between 1000 and 3500 cm−3 at 15–20 km altitude (Fig. 27b), depending on cutoff rigidity. These values are between a factor 2–3 smaller than the estimated ion concentrations assuming loss by recombination alone. Similar factors of about 3 occur at lower altitudes in unpolluted air, which indicates that the scav- enging of small ions by aerosol particles is an important loss mechanism at all tropospheric altitudes. From Eq. 3, the recombination lifetime of an ion is τ =1/αn. This implies ion lifetimes due to recombination of about 3–10 min, depending on altitude. When attachment dominates, the ion lifetime is given by τ =1/βN,and typical values are about 100 s. These lifetimes set the timescale within which processes such as ion-induced nucleation must take place if they are significant. The small ions drift vertically in the electric field created by the negatively-charged Earth and the positively-charged ionosphere. The field strength is E ∼ 100 V/m at ground level, producing an drift velocity for small ions of about 1.5 cm s−1 and a drift distance of up to about 10 m over their lifetime. At an altitude of 15 km, however, E ∼ 2 V/m due to the higher conductivity of the air. Here the drift velocity is only 0.1 cm s−1 and the drift distance is less than 1 m. Ions can be transported substantially further by winds, both vertically and horizontally. Small ions are very efficiently scavenged by cloud droplets in a similar way to the aerosol attach- ment described above. This results in a sharp reduction of the electrical conductivity, σ, inside clouds (since σ = neµ, where e is the electronic charge and µ the electrical mobility). Measured conductivities inside non-electrified clouds are reduced by a factor 10–40 relative to clear air, and by a factor 200-500 inside electrically active clouds (E =30kVm−1) [102]. Therefore, to a good approximation, almost all the charge in a cloud can be assumed to reside on the droplets. The difference between the conductiv- ities of clear air and of clouds causes a layer of space charge to form at the cloud boundary regions, which generates a relatively high vertical electric field, Ez, inside the cloud and maintains continuity of the vertical conduction current, Jz = σEz.Typical equilibrium droplet charges at cloud boundaries are quite large—about 100 e—and take about 13 min to be established by the positive and negative ions drifting into the cloud from below and above, respectively. Updrafts and downdrafts can carry these highly charged droplets and aerosols deeper inside the clouds. 4.3.5 Physical mechanisms Having summarised the general characteristics of the interaction of cosmic rays with the atmosphere, we will now consider how these interactions may influence cloud microphysics. These processes fall into three categories, as shown schematically in Fig. 29: 1. Aerosols. 2. Ice particles. 3. Cloud electricity. Each of these processes is discussed in detail below. galactic cosmic rays solar wind modulation Sun in the heliosphere tropospheric & stratospheric ions & NO/OH radicals Earth cloud aerosols ice particles electricity Fig. 29: The three categories of cloud processes that may be affected by galactic cosmic rays, whose intensity is modulated by the solar wind. 4.4 Aerosols 4.4.1 Cloud condensation nuclei Atmospheric aerosols are liquid or solid particles suspended in the air. The atmosphere contains sig- nificant concentrations of aerosols, sometimes as high as 106 cm−3. Aerosol composition varies sig- nificantly with respect to location and size distribution, with the smallest aerosols often being clusters of volatile species such as sulphuric acid and water (formed from gas-to-particle conversion) and the largest often being inorganic salts and dust particles (§4.4.3). Aerosol sizes can often be described by three quasi-distinct modes comprising a nucleation mode (diameter range ∼1–100 nm), an accumulation mode (∼0.1–1 µm) and a coarse mode (>1 µm). Many different kinds of aerosol are capable of acting as condensation nuclei (CN) but only a subset constitute cloud condensation nuclei (CCN). These can activate into cloud droplets when the relative humidity exceeds 100% (or, equivalently, when the water vapour supersaturation, S,exceeds 0%). The presence of a largely abundant supply of CCN ensures that the maximum water vapour supersaturations in the atmosphere rarely exceed values of about 1% since higher values are arrested by the removal of water vapour during droplet growth. These values are far below the supersaturations (∼500%) required to activate small ions into droplets, as in a classical Wilson cloud chamber [74]. Therefore, if GCRs can affect clouds, it is a priori likely to be through some influence on CCN. The activation process can be understood from the Kohler¨ curves (Fig. 30), which show the equilib- rium S (and therefore equilibrium vapour pressure) over droplets of various sizes and containing various masses of dissolved salts. The equilibrium S of pure water droplets increases with decreasing radius due to the effect of curvature (Kelvin’s equation; ln (p/p0) ∝ σ/r, where p/p0 is the water vapour saturation ratio, σ is the air-water surface tension and r is the droplet radius). However dissolved salts reduce the 3 equilibrium S due to a reduction of the molar concentration of the water (Raoult’s law; p/p0 ∝−1/r ). The latter effect dominates at small radii, i.e. at high solute concentrations. Recent studies [54] have also indicated the importance of organic surfactants in reducing the surface tension of cloud droplets, resulting in an increase of droplet number concentrations at lower supersaturations The droplet number density in liquid water clouds depends upon the cooling rate of the air as it enters the cloud (since this affects the peak S that is reached) and upon the concentration, size and chem- ical composition of the CCN. Although highly variable, typical number densities are a few × 100 cm−3 in continental clouds and a few × 10 cm−3 in marine clouds (Fig. 31). Number densities are usually higher in convective clouds than in stratiform clouds. Once activated, droplets grow by diffusion of water vapour. Diffusional growth is rather slow and it is unusual for droplet radii to exceed 20–30 µmbythis process. Cloud droplets typically attain sizes of 10 µm within a few minutes but take over an hour to reach 100 µm (since the growth time ∝ r2/S). Droplet collision and coalescence (which occurs when droplets collide while falling under the influence of gravity) takes over as the principal growth mechanism for radii above about 20 µm. For clouds to generate rainfall, some drops must grow to precipitable sizes of 1 mm or greater. This is achieved either by collision and coalescence of droplets or by ice formation (glaciation). Ice formation usually occurs in only a small fraction of the cloud droplets, allowing these to preferentially grow by vapour diffusion due to the lower vapour pressure of ice compared with water droplets (§4.5.1). The ability of a cloud to generate rain is an important factor in determining its lifetime. 4.4.2 Effect of CCN changes on cloud radiative properties and lifetime Cloud radiative properties: The effect that a change in the CCN number concentration has on the radiative properties of a cloud can be quantitatively estimated as follows [76, 77]. Assuming the liquid water content and depth of the cloud is fixed, then its optical thickness, τ,isgiven by τ ∝ Nr2,where N is the droplet number concentration and r the mean droplet radius. Since N ∝ r−3, this indicates τ ∝ N 1/3. Therefore a change of the droplet number concentration by ∆N leads to a change of the 0.8 0.7 293 K 0.6 pure water (Kelvin's eq.) 0.5 0.4 0.3 0.2 critical supersaturation µ for 0.1 m (NH4)2SO4 0.1 0.0 ater vapour supersaturation (%) supersaturation ater vapour W 0.05 µm 0.1 µm 0.5 µm: dry diameter -0.1 (NH4)2SO4 -0.2 solute effect (Raoult's law) -0.3 0.1 1 10 100 Droplet diameter (µm) Fig. 30: Kohler¨ curves showing the equilibrium water vapour supersaturation at 293 K for droplets of pure water (dotted curve) and for droplets containing various masses of dissolved (NH4)2SO4 (solid curves) vs. diameter of the droplet [75]. The water vapour supersaturation, S (%) =(p/p0−1)·100, where p is the partial pressure of the water vapour and p0 is the saturated vapour pressure over a plane surface of water at this temperature. In the indicated example, an ambient water vapour S of 0.15% (dashed line) exceeds the critical value for all ammonium sulphate aerosols with dry diameter ≥ 0.1 µm. These aerosols will therefore activate and grow into cloud droplets, whereas smaller aerosols remain as unactivated haze particles. Droplets below their corresponding equilibrium curve will shrink by evaporation whereas those above will grow by condensation (the indicated droplets correspond, for example, to a dry diameter of 0.05 µm). 10,000 ) Continental (Buffalo, NY) -3 Continental (Australia) 1,000 Continental (typical) Marine (Australia) 100 Atlantic Hawaii CCN concentration (cm 10 0 0.2 0.4 0.6 0.8 1 1.2 Water vapour supersaturation (%) Fig. 31: Measurements of CCN concentrations at several sites: marine (solid curves) and continental (dashed curves) as a function of the water vapour supersaturation [75]. The CCN concentrations are equal to the cloud droplet concentrations at a given supersaturation. optical thickness by ∆τ given by ∆τ 1 ∆N = · (4) τ 3 N The albedo (reflectivity), A,ofacloud is the fraction of incident radiation that is reflected into the backward hemisphere. For the scattering of solar radiation by clouds [76, 77], τ A ≈ (5) τ +6.7 Differentiating Eq. 5 and combining with Eq. 4 gives, ∆A ∆N =(1− A) · (6) A N The rather thin stratiform clouds that cover an appreciable fraction of the Earth’s surface, and especially marine regions, have an albedo of about 0.5 and a droplet number concentration of about 100 cm−3 or less (Fig. 31). Equation 6 shows that these clouds are very sensitive to changes in the CCN number concentration; their reflectance changes by 0.5% or more per single additional cloud droplet per cubic centimetre of air! This in turn indicates that GCR-induced changes in the CCN number concentration of only a few per cent could produce significant effects on the radiative properties of such clouds. As a numerical example, Figure 32 shows the variation of cloud reflectivity with cloud depth and cloud droplet number density for a fixed liquid water content of 0.3 g m−3 (most data confirm that there is little or no dependence of liquid water content on cloud droplet number density). 1.0 cloud thickness (m) 0.8 1500 0.6 500 0.4 Reflectivity 100 0.2 50 0 10 100 1000 Cloud droplet number concentration (cm-3 ) Fig. 32: The variation of cloud albedo with cloud thickness and droplet number concentration for a fixed liquid water content of 0.3 g m−3 [78]. Cloud lifetime: The second effect of an increase in CCN concentration is to suppress rainfall and thereby to increase cloud lifetime. This has been observed over oceans in ship tracks [79] and, recently, also over land [80]. The latter study used NOAA satellite data to investigate clouds that formed down- wind of industrial sites located in pristine areas. The otherwise uniform cloud data from these regions was streaked with bright (highly reflective) cloud plumes from the industrial sites. The droplets in these plumes were found to be more numerous than the nearby regions and of a smaller diameter—typically less than 10 µm and therefore below the threshold size for them to coalesce efficiently and precipitate. In contrast the droplets outside the plumes measured more than 25 µmindiameter. The high reflectivity of the plumes resulted from the high droplet number density at fixed liquid water content (Fig. 32). Inde- pendent analysis of data from the Tropical Rainfall Measuring Mission confirmed that these plumes did indeed produce less rain and therefore had a longer lifetime than clouds in the nearby regions. secondary aerosol sources primary gaseous SO /DMS emissions NO emissions 2 x organic emissions GCR GCR gas-phase gas-phase gas-phase photochemistry photochemistry photochemistry condensable H2SO4 HNO3 organic vapours primary H SO 2 4 NH emissions emissions 3 gas-to-particle GCR nucleation H2O vapour aerosol particles primary particulate primary particulate sea salt inorganic emissions organic emissions (dust, carbon, etc.) primary aerosol sources Fig. 33: The main sources of atmospheric aerosols. These are classified as primary if they are injected directly into the air or secondary if they result from gas-to-particle conversion in the atmosphere. The processes that may be affected by galactic cosmic rays (GCRs) are indicated by wavy arrows. 4.4.3 Production and loss of atmospheric aerosols The main sources of atmospheric aerosols are summarised in Fig. 33. Aerosols are classified as either primary or secondary, and they may be of either natural or anthropogenic origin. Primary aerosols are those injected directly into the air (e.g. by wind erosion, sea spray, pollen, etc.). Secondary aerosols are those created by gas-to-particle nucleation of vapour molecules. Inorganic aerosols are usually weakly acidic, with the most common aqueous cationic components being H+, ammonium and sodium, and with common anionic components being sulphate, chloride and nitrate. Such aerosols are hygroscopic. Aerosols can also be partly or wholly composed of organic compounds derived from plant waxes and combustion sources. These aerosols may be either hygrophobic or hygroscopic. Secondary aerosols may originate from emissions of non-condensable vapours followed by gas-phase chemical conversion into condensable (i.e. low vapour pressure) aerosol precursors, e.g. SO2 oxidation to H2SO4. Clouds are an important source as well as sink of aerosols since they provide efficient sites for the scavenging and chemical processing of aerosols and aerosol precursors. These pathways for aerosol production are shown in Fig. 34. Once formed, aerosols have both a direct and an indirect effect of the climate by their influence on radiative forcing. The direct effect is due to scattering and absorption of the incoming shortwave radiation. Absorptive aerosols such as black carbon have a net positive forcing (warming) and reflective ones such as hygroscopic aerosols have a net negative forcing. The indirect effect is due to a change Sun Indirect radiative forcing direct radiative forcing aerosol cloud lifetime cloud albedo albedo cloud ice cloud liquid particles droplets ice nuclei CCN rainfall GCR aerosol particles aerosol wet and dry aerosol production aerosol production deposition in clouds emissions in the air GCR aerosol precursors gaseous emissions Fig. 34: Aerosol production and loss in the atmosphere, and the effects on clouds and climate of changes in the aerosol number concentration. The processes that may be affected by galactic cosmic rays (GCRs) are indicated by wavy arrows. in the number concentration of CCN and, consequently, to changes in cloud reflectivity and lifetime (§4.4.2). A small subset (about 10−6)ofCCN constitute ice nuclei and these have a strong influence on cloud radiative properties and lifetime (§4.5.1). Aerosols are removed from the atmosphere by wet or dry sedimentation (Fig. 34). The residence time depends on their composition, size and geographical location. The lifetime for wet removal is about 8days in the lower troposphere and about 3 weeks in the middle to upper troposphere. This applies to the accumulation-mode aerosols, which constitute the predominant CCN type. Larger aerosols have a shorter residence time due to large settling velocities. Smaller aerosols are removed relatively rapidly by coagulation, which transforms them into fewer, larger aerosols. A consequence of the short lifetime of tropospheric aerosols is a large spatial variation of their composition, size and number concentrations. 4.4.4 GCR-aerosol interactions If increase of the GCR flux were to translate into increases in CCN number concentration then this would extend cloud lifetimes, consistent with the satellite data (Fig. 2). There are several processes in the production and loss of aerosols that may be affected by GCRs, as indicated in Figs. 33 and 34. An important source of new aerosol particles in the atmosphere is the nucleation of ultrafine con- densation nuclei (UCNs) from precursor vapours, of both natural and anthropogenic origin. Despite intensive research over several decades, the origin of the ubiquitous background of ultrafine aerosols in the troposphere has not yet been determined. Moreover even the fundamental mechanism that leads to new particle formation remains poorly understood. Understanding these processes is crucial to deter- mining the contributions of both natural and anthropogenic aerosol effects on radiative forcing of the climate. It has been suggested that ionisation from GCRs may play a key role in the formation of new aerosol particles [81]–[88]. An important precursor vapour for UCN and CCN is sulphuric acid. However the classical theory of binary H2SO4–H2O homogeneous nucleation fails to explain observations of new ultrafine particle formation in clean regions of the lower atmosphere, such as occurs over oceans and in pristine continental air [89]–[92]. Typically the nucleation rates predicted by classical theory are far lower (by as much as 10 orders of magnitude) than the experimentally-observed rates. Recent modelling work [93]–[95] demonstrates that thermodynamically-stable charged clusters, caused by vapours condensing onto ions, can form at much lower ambient vapour concentrations and grow significantly faster than neutral clusters. The steps involved in the creation of CCN from condensable vapours (in this case, sulphuric acid) are shown in Fig. 35. Molecular H2SO4-H2O clusters form and evaporate continually by kinetic motion. Under suitable conditions some clusters will reach the critical size of about 1–2 nm diameter. Once the critical size is reached, continued growth of the cluster becomes preferential thermodynamically. The nucleation of aerosols in the atmosphere involves several competing processes which include molecular clustering, evaporation, scavenging of condensable vapour by pre-existing aerosols, and sedimentation by rainfall. In this environment, electrically charged embryos have a competitive advantage over neutral embryos. Charged clusters provide additional electrostatic attractive forces with polar molecules, allow- ing critical embryos to form with fewer molecules than for neutral clusters. Therefore ions can greatly enhance the rate of formation of new particles in regions where the concentration of condensable vapours is too low for stable neutral clusters to form at an appreciable rate, such as frequently occurs in the marine boundary layer. The key parameters controlling the rate of new particle formation are the concentrations of condensable vapours, the GCR ionisation rate, and the surface area of pre-existing aerosols. Once formed, the UCN continue to grow. The main growth process up to diameters of about 10 nm is molecular condensation; for larger sizes the main growth mechanism is coagulation of existing CN. During the growth of CN to CCN, other vapours in the atmosphere such as ammonia, nitric acid and low volatility organic compounds are known to be important. The growth rate is expected to be enhanced by the presence of ions, becoming less significant with increasing size of the aerosol particle. cloud droplets CCN (~10-20 µm) CN (~100 nm) + condensation activation – + + + & coagulation critical embryos (~1-2 nm) con densation GCR H2O vapour + condensation – cluster + nucleation H2SO4 evaporation vapour UCN sub-critical GCR embryos (<1-2 nm) S02 DMS Fig. 35: The nucleation of ultrafine condensation nuclei (UCN) from trace sulphuric acid vapour, followed by aerosol growth into condensation nuclei (CN) and cloud condensation nuclei (CCN), which can activate into cloud droplets. The precursor of sulphuric acid is SO2, produced anthropogenically or, in remote marine environments, predominantly from dimethyl sulphide (DMS) from plankton. The processes that may be affected by galactic cosmic rays (GCRs) are indicated by wavy arrows. Charged aerosols are expected to have an enhanced growth rate and reduced evaporation relative to neutral aerosols. GCRs may also affect the activation of CCN in droplets. This enhancement is largest when the two colliding particles have opposite sign (+−), but there is also an increased rate between one charged and one neutral particle (+0 or −0) due to image charge attractions, when compared with two neutral particles (00). Finally the activation of CCN into cloud droplets may also be influenced by charge. However the effect is expected to be small for the typical electric charges found on aerosol particles under fair-weather conditions. There is a continual interchange between charged and neutral particles as small ions diffuse onto existing CN and CCN, either neutralising or charging them in the process. This implies that ions may po- tentially affect the production rate of a large fraction of the CCN produced by gas-to-particle conversion, regardless of whether or not the original UCN were produced via ion-mediated processes. Aerosol particles and trace vapours are continually being scavenged from the atmosphere by rain- fall. Following a precipitation event, the air is left with a relatively low aerosol particle number concen- tration and a low aerosol surface area. It is under these conditions that the nucleation of fresh UCN is most likely to occur, provided sufficient concentrations of precursor vapours are present. Furthermore, for clean environments, this indicates that the rate at which new particles are produced and grow into CCN can strongly affect the lifetime and radiative properties of the clouds in these regions, since there is rarely sufficient time for a large CCN population to form. Radon activity in reaction vessel [pCi m ] 0 100 200 300 400 500 600 6 14 10 a) b) ] 220 -3 ] 0.1 Gy doses of 86 Rn 12 -3 cm 4 10 105 8 N2:O2 = 4:1 SO2 = 300 ppb C2H4 = 100 ppb 6 O3 = 160 ppb 104 4 600 s exposure filtered (aerosol-free) average natural Aerosol concentration [cm Aerosol concentration Rn activity Paris air GCR ionisation equivalent 2 Aerosol concentration [x 10 at ground level 103 0 0 0.5 2.5 3.0 3.5 4.0 4.5 03691215 18 Radon activity in reaction vessel [Bq m-3] Time [h] Fig. 36: Experimental evidence for enhanced nucleation of aerosols from trace gases caused by ionising radiation (α particles from radon). The measurements involve a) filtered Paris air with high irradiation doses [97] and b) artificial air with high trace gas concentrations but at naturally-occurring radiation doses [98]. 4.4.5 Experimental knowledge of ion-aerosol interactions There are only sparse experimental data on the effect of ions in the atmosphere on new particle formation —and none, to our knowledge, on the effect of ions on particle growth from CN to CCN, or on the activation of CCN into cloud droplets. Observations have been made of nucleation bursts of CN in the atmosphere that cannot be explained by classical theories. For example, Horrak˜ et al. [96] reported the spontaneous formation of bursts of intermediate size ions in urban air, which they suggest may be due to ion-induced nucleation. Also, Clarke et al. [89] observed formation of new ultrafine particles in the 7 tropical marine (Pacific) boundary layer that could not be explained by classical binary (H2SO4-H2O) homogeneous nucleation theory at the measured low ambient concentrations of sulphuric acid (1–5·107 molecules cm−3). However a recent study by Yu and Turco [93, 95] based on an ion-mediated model is able to repro- duce the observations of Clarke et al.. Their model indicates that the nucleation rate of fresh CN in the marine boundary layer is generally limited by the available ion production rate from GCRs. In contrast, the nucleation rate in the upper atmosphere is generally limited by the trace vapour concentration since the temperatures are lower and the trace vapour saturation ratios correspondingly higher, and so binary homogeneous nucleation can occur at an appreciable rate. This provides a possible reason why the solar modulation signal only appears in clouds below about 3 km. Direct experimental evidence that ions are involved in the nucleation of new particles under at- mospheric conditions is lacking. However positive effects with ions have been seen. Two examples are shown in Figs. 36. Bricard et al. [97] observed new particle production in filtered (aerosol-free) Paris air exposed to very high radiation doses (3 · 108 Bq m−3 × 300 s). On the other hand, Vohra et al. [98] carried out experiments with radon at naturally-occurring ionisation levels of 3–15 Bq m−3 and ob- served new particle production proportional to ionisation rate, but they used artificial air containing high concentrations of trace gases (300 ppb SO2, 100 ppb C2H4 and 160 ppb O3). 7The boundary layer is the layer of the atmosphere within about 1 km of the Earth’s surface, within which air is subject to turbulence, friction effects and surface heating. The region extending from about 1 km to the tropopause is known as the free troposphere. 4.5 Ice particles 4.5.1 Overview The second class of processes by which GCRs may affect clouds concerns ice particles. The formation of ice in clouds is important for several reasons: 1. In mixed-phase clouds, ice particles grow rapidly at the expense of liquid water droplets. When the water vapour supersaturation relative to liquid water is 0%, the supersaturation relative to ice is much higher, as large as 50% (Fig. 37). Since supersaturation is the ‘driving force’ that determines growth rate, the ice particles grow rapidly, reducing the ambient supersaturation and causing the liquid water droplets to evaporate. The ice particles rapidly grow to a size where they sediment and ‘rain-out’ the clouds below.8 2. The freezing of supercooled water releases latent heat, which affects cloud dynamics. 3. Ice particles modify the radiative properties of clouds, both by increasing sedimentation and by changing the reflectivity (particles that are larger and crystalline). As an indication of the impor- tance of ice particles, the IPCC finds that uncertainties in the fraction of frozen water in clouds results in differences of up to 17 Wm−2 in their estimates of globally-averaged cloud forcing [49]. 50 40 30 20 10 0% SS relative to liquid water Supersaturation relative to ice (%) relative Supersaturation 0 0 -10 -20 -30 -40 Temperature (oC) Fig. 37: The supersaturation relative to ice for water vapour that is in equilibrium with liquid water at the temperature indicated on the x axis. In the temperature range from 0◦Cto-40◦C liquid water can exist in a supercooled state in the absence of an ice nucleus. Below -40◦C, water freezes homogeneously, i.e. without need of a distinct ice nucleus. Once a cloud extends to altitudes where the temperature is below 0◦C, ice crystals may form. Two phase transitions can lead to ice formation (Fig. 38): (a) the direct deposition (sublimation) of water vapour to ice and (b and c) the freezing of a supercooled liquid droplet. The latter may occur either by the transformation of a supercooled liquid droplet into an ice particle (freezing nucleation) or collision of a supercooled liquid droplet with an ice nuclei (contact nucleation). The relative importance of these three freezing modes has not yet been established. The freezing may proceed either on a suitable ice nu- cleus (IN) (heterogeneous nucleation), as shown in Fig. 38, or else occur with pure water (homogeneous nucleation). For homogeneous nucleation to take place, a statistical fluctuation of the water molecules must occur to produce a stable, ice-like structure that can serve as an ice nucleus. For typical cloud droplet dimensions (∼10 µm), homogeneous nucleation occurs at about -40◦C. Therefore, for clouds in the temperature range between 0◦C and -40◦C, ice particle nucleation is heterogeneous and requires a suitable IN. Field measurements show that clouds contains a great deal of supercooled liquid water in this temperature range, since ice nuclei are very rare in the atmosphere. The IN number concentration is 8This is the principle behind cloud-seeding, in which a suitable ice nucleus material, e.g. AgI, induces the freezing of supercooled cloud droplets. b) freezing nucleation GCR supercooled liquid droplet a) deposition nucleation ice nucleus ice particle supercooled liquid droplet c) contact nucleation Fig. 38: Processes for ice particle formation in clouds, involving a) deposition nucleation: the direct sublimation of water vapour to the solid phase on an ice nucleus, b) freezing nucleation: condensation of a supercooled liquid droplet on a suitable ice nucleus, followed by freezing as the temperature falls, and c) contact nucleation: the freezing of a supercooled liquid droplet (already formed on a CCN) by external contact with an ice nucleus. GCRs may affect the efficiency of CCNs to act as ice nuclei and may directly affect freezing nucleation. typically ∼1 /litre at about -20◦C, increasing by a factor 10 for each 4◦Cofadditional cooling. This may be compared with a CCN number concentration of ∼ 106 /litre or, said another way, only one in a million CCN constitutes a suitable IN at -20◦C. Identifying such particles is a difficult task, and IN remain poorly understood. However efficient IN are generally insoluble in water and have a chemical bonding and crystallographic structure similar to ice. Examples include various insoluble salts (such as AgI), certain clay particles, various organic materials and even bacteria (which is surprising since neither of the last two have a crystalline structure). Although some IN have been identified, the number of ice particles found in clouds often exceeds the measured IN concentration by several orders of magnitude. Part of this discrepancy can be attributed to ice multiplication in secondary processes. When a droplet freezes at temperature between about -5◦C and -10◦C, mechanical stresses lead to the ejection of small ice fragments, which in turn act as efficient ice nuclei. Collisions between dense graupel particles and fragile dendritic crystals also generate ice splinters. However the understanding of these processes is also poor. Recent measurements [99] suggest that freezing-thawing cycles may be an important factor in IN production. These authors found that the thermal history of droplets affects the temperatures at which they eventually freeze. In summary many observations of ice particle in clouds cannot be explained quantitatively and, in particular, there appears to be a great lack of IN to account for the observed numbers of ice particles in clouds. 4.5.2 GCR-ice particle interactions Enhanced heterogeneous ice nucleation by electrification has been proposed by several workers. For example, Tinsley [100, 101] has proposed that cosmic rays have an important influence on cloud micro- physics and climate through the following sequence of events. Cosmic rays generate ionisation in the atmosphere and determine the magnitude of the vertical conduction current, Jz (§4.6.2). This current generates highly charged droplets (∼>100 e)atthe upper and lower boundaries of clouds due to the ac- cumulation of space charge (§4.3.4). When these droplets evaporate they leave behind highly charged aerosols which are coated with extra sulphates and insoluble organic compounds scavenged while the droplet existed. (It has been shown that neither electric charge nor aerosol material is not lost when a droplet evaporates.) These highly charged and coated “evaporation nuclei” constitute efficient ice nuclei, either by deposition nucleation or by contact nucleation. The presence of charge enhances collisions of the evaporation nuclei with other liquid droplets by “electroscavenging” (§4.6.2), thereby generating ice particles in clouds. If this sequence of events is correct, it would imply that increased GCR intensity leads to in- creased ice particle formation in clouds, which in turn releases latent heat, increases cyclone activity, and increases rainfall. Tinsley supports his claim with a study of the Vorticity Area Index (a measure of regional-scale cyclone motion) which he shows to decrease during Forbush decreases of the GCR flux. During Forbush events, which are caused by severe solar disturbances (CMEs), the GCR intensity is reduced by around 3–10% over a period of about 1 day, with a recovery time of a few days. The key uncertainty in this sequence of events is whether or not charged aerosols, perhaps together with cloud processing, are more effective as IN. An enhancement due to charge is supported by very little experimental work so far [102]. Some experiments, but not all, have reported positive effects. For exam- ple, cloud chamber experiments at the University of Missouri-Rolla in 1980 [103] found an enhancement of frozen droplets at -33◦Cinregions where cosmic rays traversed the cloud chamber. The presence of ions was observed to raise the threshold temperature for homogeneous ice nucleation by about 2 K. The cloud-processing and evaporation aspects of Tinsley’s scheme are qualitatively supported by recent stud- ies [99] which indicate that the morphology of any crystallised solid in an aerosol strongly influences its effectiveness as an IN. In summary, there seem to be some indications that ionising radiation may affect ice nucleation, but the experimental picture is far from clear. 4.5.3 Polar stratospheric clouds Polar stratospheric clouds (PSCs), also known as mother-of-pearl or nacreous clouds, play a key role in the process of ozone depletion in the polar regions, especially Antarctica. PSCs are clouds that form in the cold polar stratospheric winters where, despite the dryness of the stratosphere, the temperature drops low enough for condensation and freezing to occur. At temperatures above the ice frost point the particles may be either liquid solutions of nitric acid, sulphuric acid and water, or else solid nitric acid and ice in the molar ratio 1:3—so called nitric acid trihydrate (NAT). In other parts of the world the stratosphere is too warm for these clouds to form, which is one reason why the ‘ozone hole’ is confined to the Antarctic region. Normally, the chlorine of anthropogenic ozone-depleting chemicals is locked up in relatively inert and stable chlorine compounds. However, during the Antarctic winter months (June to August) when the region receives no sunlight and is isolated by a wind circulation pattern known as the polar vortex, the stratosphere becomes cold enough (190–195 K) for PSCs to form. The PSCs provide a heterogeneous catalytic surface on which chlorine can be converted from inert ‘reservoir’ species, such as ClONO2 and HCl, into active species: HCl(s) + ClONO2 → Cl2 + HNO3(s) (7) In the presence of sunlight the Cl2 photolyses, producing free Cl atoms which react with the ozone, thus destroying the ozone layer. Since the reaction requires sunlight, it only begins when the sunlight returns in the Antarctic spring (September to October), before the PSCs have had a chance to evaporate. The ozone hole disappears again when the Antarctic air warms up enough during late spring and summer In reaction 7, the Cl2 is released but the HNO3 remains in the PSC particles. Since gaseous HNO3 can convert active chlorine to reservoir species, this further facilitates ozone destruction. In fact, massive ozone depletion requires the abundance of gaseous HNO3 be very low. In principle this nitric acid would be liberated when the PSCs evaporated with the return of the Sun. However, freezing of PSC particles allows the selective growth of a small number of particles that subsequently become large enough to sediment out of the stratosphere (§4.5.1). This process leads to denitrification of the polar stratosphere and to a strongly enhanced ozone loss. Several mechanisms are recognised to be important for the formation of solid polar stratospheric cloud particles. However, these persistent and optically very thin ice clouds cannot be explained by any recognised mechanism. Since PSCs exist on a very large scale, often over several thousand kilometre regions with little spatial variability, the possible mechanisms are tightly constrained. Laboratory ex- periments have excluded the possibility that these solid particles form by crystallisation of the liquid aerosols due simply to large scale cooling. An intriguing suggestion—so far unexplored—is that these clouds form by deposition nucleation of nitric acid and water directly onto cosmic ray-generated ions or ion clusters (process ‘a’ in Fig. 38). However, to date, there have been no experiments that can confirm or dispute this possibility. An understanding of the freezing mechanism of PSCs appears to be critical to a complete understanding of denitrification and ozone loss [104]. 4.6 Cloud electricity 4.6.1 Overview The third class of processes by which GCRs may affect clouds concerns the electrical nature of the atmosphere. Except for a contribution from radioactive isotopes near the land surface (Fig. 28), GCRs are responsible for generating all the fair-weather atmospheric ionization between ground level and the mid mesosphere, at about 65 km altitude (§4.3.3–4.3.4). As such, GCRs fundamentally underpin the global electrical circuit. mesosphere ~60-80 km altitude (ionosphere) +250 kV 105 Ω (fair weather value) 1250 A fair weather current, ~10 km altitude J ~ 2.4x10-12 Am-2 thunderstorm generators (40 lightning -1 flashes s ) 200 Ω 0.7 F (τ = RC 104 Ω ~ 2 min) (fair weather value) 0 V Earth's surface Fig. 39: Schematic of the global electrical circuit. The current generator is thunderstorms, which are predomi- nantly located in the tropics, and the return path is the global fair-weather current flowing between the ionosphere and ground. GCRs play a central role in these processes. The atmospheric electric circuit (Fig. 39) involves a global current of 1250 A which is sustained by thunderstorms continuously active around the tropics. The Maxwell current density below thunderstorms is the sum of several components, of which lightning contributes about half, with the remainder from electrical conduction, air convection, and precipitation. The thunderstorms carry negative charge to the ground and an equivalent positive current flows up to the ionosphere. Due to the high currents, large electric fields are generated above thunderstorms and the air is likely to break down electrically. Indeed, during the last decade, optical flashes known as sprites and elves have been detected above thunderstorms carrying the positive current up to the ionosphere. This current generator maintains the ionosphere at a relative positive potential of about 250 kV. Since the ionospheric potential drives a fair weather current of 1250 A, it represents a very powerful continuous generator of about 300 MW. The return current between the ionosphere and the Earth’s surface flows throughout the atmosphere, in regions of disturbed and undisturbed weather, and is carried by vertical drift of small ions. The average fair-weather current density, J =2.4 pA m−2. Electric fields present in the atmosphere vary between fair weather values of typically 100 Vm−1 at the surface and about 2 Vm−1 at 15 km altitude (due to the higher air conductivity). In clouds, the electric fields are generally ∼<500 Vm−1 butreach about 100 kVm−1 in thunderstorms before a lightning discharge (Table 3). Table 3: Typical maximum electric fields measured inside clouds [105]. Electric fields in all types of non- thunderstorm clouds are generally ∼<0.5 kVm−1.For comparison, the clear-air electric field is about 0.1 kVm−1 at ground level. The dielectric field strength of dry air at stp is 30 MVm−1, which represents the breakdown field strength across plane parallel electrodes. The threshold electric fields for wet air breakdown are about 1–2 MVm−1. Cloud type Maximum electric field (kVm−1) stratus 1 stratocumulus 1.5 cirrostratus 1.5 altostratus 5 nimbostratus 15 thunderstorm 100 2.0 12 a) Polar region b) USA ) 0.66 30 ) -1 /year) s -1 GCR 6 s -2 ) lightning 11 -2 -2 1.6 25 0.62 Am 20 10 -12 1.2 0.58 (8 km altitude) (x 10 atmospheric 15 9 GCR (in 2-10 km column) current GCR intensity (cm GCR ion pair rate (cm GCR ion pair rate Atmospheric current density, J 0.8 0.54 Lightning frequency (x10 10 8 1965 1970 1975 1980 1985 1988 1992 1996 2000 Year Year Fig. 40: a) Solar-cycle variation of the atmospheric current density, J, and the GCR intensity, both in the polar region, [107] and b) frequency of lightning recorded in the United Sates and change of GCR intensity for 1988– 1999 [108]. 4.6.2 GCR - cloud-electricity interactions Global electrical circuit: By their effect on the ion pair concentration and, perhaps, on lightning fre- quency, GCRs can in principle affect the atmospheric conductivity, σ, the ionospheric potential, the atmospheric current density, J, and the atmospheric electric field, Ez.Astudy of the variation of atmo- spheric current density in the polar region over the period 1965–1985 (Fig. 40a) [107] shows evidence for a solar modulation. The increase of J around the minimum of the solar cycle is consistent with the increased conductivity of the air due to the higher GCR intensity. However, there is a second inference to be made: there must be a simultaneous increase of the current source, i.e. of lightning frequency, in order to sustain the higher net flow of fair-weather charge between the ionosphere and the ground. Some direct evidence that supports this is seen in Fig. 40b) [108]. This suggests that the efficiency of charge separation in thunder clouds may be influenced by the ionisation concentration from GCRs (a mechanism that is consistent with this picture is discussed below). However other studies of a possible relationship between solar activity and the frequency of thunderstorms or lightning have been made, and the results are not conclusive with some reporting a positive correlation, some a negative correlation and some none of any significance (see, for example ref. [106] and references therein). Therefore more experimental data are required to clarify the situation. Since lightning is the dominant source of NOx in the marine troposphere, if GCRs do indeed affect lightning frequency, it would imply they also modulate the production of this important reactive trace radical in the troposphere [109]. Note that this production mechanism for NOx (and OH radicals) occurs both below thunderclouds (due to lightning) and above them (due to sprites and elves; § 4.6.1). Rainfall: Stozhkov et al. have studied the influence of Forbush decreases of GCR intensity on precip- itation in Brazil and in the former Soviet Union [55, 110]. The combined experimental data for around 70 Forbush events (Fig. 41a) indicate a decrease of rainfall of (17 ± 3)%onthe day of the Forbush de- crease. These authors have also studied ground-level solar proton events (solar cosmic rays), where the cosmic ray intensity increases for a period of about a day due to high energy solar events (Fig. 41b). Here too they find some evidence for an effect, (13 ± 5)% (53 combined events), and with a consistent sign, namely an increase of rainfall. In a separate study of Forbush events, Pudovkin and Veretenenko have reported [111] a correlated short-term decrease of cloudiness using data obtained visually in a narrow range of latitudes (60N–64N), which is consistent with the global satellite observations (§2). a) Change in rainfall (%) 10 -35 -25 -15 -5 5 15 25Day 35 0 -10 Day 0 = Forbush onset -20 Change in rainfall (%) b) 15 10 5 -30 -20 -10 10 20Day 30 0 -5 -10 Day 0 = SCR event -15 Fig. 41: Variation of rainfall recorded in the former Soviet Union and Brazil during short-term changes the cosmic ray intensity [55]. The relative change in rainfall in the days spanning a) decreases of the cosmic ray intensity due to Forbush events (70 combined events) and b) increases of the cosmic ray intensity due to ground-level solar cosmic ray events (53 combined events). These data suggest that GCR ionisation may affect the precipitation efficiency of clouds. One candidate physical mechanism is an enhanced formation of ice particles (§4.5). Another is the increased efficiency of droplet growth by collision and coalescence, which is the dominant process by which cloud droplets grow in size from cloud droplets (about 20 µm diameter) into raindrops (about 1 mm). Cloud droplets become charged by the diffusion of small ions. Measurements indicate that the mean electronic charge on a non-thunderstorm cloud droplet is approximately represented by q(e)=4.1 r2(µm2), where r is the droplet radius [102]. In thunderstorm conditions, the charge is about an order of magnitude larger, q ∼ 40 r2. The presence of charge is expected to produce large increases of coalescence efficiency due to image charge forces, which are always attractive at sufficiently close distances (see, for example, Fig. 42). This process has been termed “electroscavenging” [100, 101]. 0.1 a) q (e) = 40 r2 (µm2) µ r1=42 m droplet 0.01 coalescence efficiency b) q (e) = 4 r2 (µm2) r1 100% RH 0.001 Collision efficiency 95% r2 + 75% 0.0001 0.5 1 2 3 4 µ Particle radius, r2 ( m) Fig. 42: The effect of charge on the collision (‘sticking’) efficiency for a falling large droplet (radius 42 µm) and smaller droplets of various radii as indicated on the x axis [105]. The droplet charges correspond to a) thunderstorm clouds (dashed lines, and b) non-thunderstorm clouds. Charge separation in clouds: Many attempts have been made to explain how charge is separated in electrically active clouds, but the actual physical mechanism remains unclear [105]. One hypothesis is that charge separation is due to a preferential activation of negatively charged CCN when a rising air parcel first exceeds the threshold supersaturation of water vapour [55]. A sign preference was first noticed by C.T.R. Wilson in 1899 in his original development of the cloud chamber, but this was for small ions and at very high supersaturations of several hundred per cent. The sign preference has been attributed to surface orientation of dipolar molecules, and a theoretical understanding has been proposed [112]. In the atmosphere, the ionisation created by GCRs is efficiently scavenged by aerosols, so many have a net charge of typically a few electronic units. If the sign preference noted above also occurs under natural conditions, then the first particles to activate into cloud droplets would be preferentially of negative sign. These negatively-charge droplets could then in principle grow to a sufficient size before the positively-charged CCN started to activate that the charges could become separated gravitationally. This would result in a region of net negative charge lying below a region of net positive charge, as observed in the lower region of typical thunder clouds. Lightning trigger: As well as providing the initial source of charge, and perhaps also the mechanism for charge separation, GCRs may play a decisive role in triggering lightning. In this case, the trigger is probably a very high energy primary, near 1015 eV [113]. These occur at a rate of about 2 km−2s−1, and they are, of course, totally unaffected by solar modulation. A typical lightning flash from cloud to ground is observed to occur as a series of leaders proceeding in rapid steps each of order 10 m length, with a pause of about 50 µs between each [105]. When the stepped leaders reach the ground, the main flash occurs as a return stroke of a few hundred amps current. The stepped leaders take about 1 ms to descend about 100 m whereas the return stroke takes about 1 µs (at one third the speed of light) to travel back along the full ionisation channel created by the leaders. The return stroke continues to bring negative current down from the cloud to the ground for between 10 µs and 10 ms, corresponding to peak currents of about 10 kA and an average charge of about 20 C. A mechanism termed ‘runaway breakdown’ has been proposed to explain the initial streamer from which the stepped leaders propagate [113]. In the core of a very high energy GCR (> 1015 eV), it is proposed that sufficient electron density exists for avalanche multiplication to occur in electric fields of 100–200 kVm−1, which are typical of thunderstorm clouds (Table 3) but well below the threshold fields for conventional damp air breakdown (1–2 MVm−1). The avalanche polarizes the local plasma and thereby increases the electric field in a positive feedback. This raises the electron density to the point where a short streamer forms, which then triggers the first leader. A distinctive signature of the runaway breakdown mechanism is the appearance of X rays of energy around 50–100 keV starting with the first stepped leader, around 1 ms before the lightning discharge, which are caused by electron bremsstrahlung in the highly conductive plasma (long mean free path). Such X rays have recently been observed in lightning storms [114, 115]. Finally we comment that the association of an ultra high energy GCR shower with a lightning discharge may help explain the phenomena of ‘rain gushes’. The latter are the familiar dramatic increase in precipitation arriving at the ground shortly after nearby lightning [105]. It is possible that rain gushes result from a sudden increase of droplet coalescence efficiency due to a sharp rise of ionisation, as described above (Fig. 42). Various attempts have been made to explain how the ionisation generated by a lightning flash could rapidly diffuse laterally over a sufficiently large area, but none is convincing. However, since the transverse width of an ultra high energy GCR shower is about 100 m, then the combination of the initial ionisation, a triggered lightning flash, X ray emission and absorption, and some mild electrical activity over the full width of the GCR shower may together generate sufficient ionisation to cause the rain gush over a ∼100 m-wide region. 5 THE CLOUD FACILITY 5.1 Concept The basic concept of the CLOUD experiment (Figs. 43 and 44) is to investigate the microphysics of GCR- cloud interactions under laboratory conditions where all the experimental parameters can be precisely controlled and measured. A beam from a particle accelerator provides an artificial source of ‘cosmic rays’ that is adjustable and precisely known. The beam illuminates a 0.5 m expansion cloud chamber anda2mreactor chamber. The chambers can be operated at any temperature and pressure in the troposphere and stratosphere. The cloud chamber simulates the conditions inside clouds, throughout the atmosphere. The reactor chamber is important for experiments involving long beam exposures of a day or more, and provides the reacted gas/aerosol samples for analysis by the external instrumentation—mass spectrometers, particle sizers, etc. Data will be recorded under a wide variety of operating conditions, and the results compared at different beam intensities, including beam-off. In this way an unambiguous study of the effects of relativistic ionising particles on cloud microphysics can be carried out. Such measurements are difficult to perform with cosmic rays in the atmosphere since the natural intensity variations are modest and—with the exception of relatively rare Forbush decreases—they follow the slow 11-year solar cycle. Furthermore, a laboratory experiment allows full control of the gas and aerosol mixture under test and, in addition, complete physical and chemical analysis of the products before, during and after beam exposure. This will greatly facilitate an understanding of the microphysics and chemistry of any effects that are observed. The challenges of a laboratory experiment are to duplicate the atmospheric conditions realistically and to ensure that the detector dimensions are sufficiently large that wall effects do not influence the measurements. Surprisingly, cloud chambers have never been operated at atmospheric conditions in a particle beam. C.T.R. Wilson was inspired to develop the cloud chamber [74, 116] after observing meteorological phenomena on the mountain of Ben Nevis in 1894. He developed the cloud chamber to try to reproduce clouds in the laboratory. Although his expansion cloud chamber was crucial to the development of nuclear and particle physics in the first half of the 20th century, and earned him the 1927 Nobel Prize in Physics, Wilson remained fascinated by atmospheric phenomena throughout his life. Indeed he devoted the latter part of his research life to seeking a connection between cosmic rays and clouds, and so it is perhaps a fitting tribute that a cloud chamber be proposed for the present studies. illumination/ UV lamps reactor chamber cloud chamber field cage liquid cooling pipes vacuum layer µ beam platform fan liquid fluorocarbon piston hydraulic storage system 01m vacuum system cloud Fig. 43: Vertical section through the CLOUD facility showing the 0.5 m cloud chamber and 2 m reactor chamber. The beam counters are not shown. 5.2 Design considerations 5.2.1 Choice of expansion cloud chamber There are several instruments capable of activating cloud condensation nuclei into droplets at the low wa- ter vapour supersaturations (few × 0.1%) found in clouds. Traditionally these are based on the diffusion cloud chamber or some variant thereof. All produce the necessary supersaturations over relatively small volumes and have inherent limitations in precision and in range of supersaturation (or, equivalently, in range of CCN size) [117]. Expansion cloud chambers, on the other hand, can in principle produce a pre- cise and uniform supersaturation over a large volume and, moreover, maintain or dynamically adjust the supersaturation over long periods. They are therefore uniquely suited to the CLOUD experiments, many of which require long times for beam exposure, aerosol growth and droplet observation (§5.5.1). Expan- sion cloud chambers can also cover the full range of supersaturations between cloud conditions and those required (∼500%) to activate small ions and embryonic aerosols in the nanometre size range. The phrase “in principle” is used above since the technical requirements are quite challenging (see below) and, to our knowledge, a cloud chamber with the performance proposed has not previously been built. However, the design requirements can be achieved with current technology, and extensive experience with cloud chambers has shown that they are high precision instruments. In particular, previous measurements by members of the CLOUD collaboration have demonstrated that the thermodynamic conditions after an Fig. 44: Cut-away view of the CLOUD facility, showing the 0.5 m cloud chamber (left) and 2 m reactor chamber (right). The external instrumentation (mass spectrometers, ion mobility spectrometers, particle sizers, etc.) is not shown. expansion are precisely known and reproducible provided the initial conditions are well-known and the expansion ratio (pressure change) is well-measured [118, 119]. 5.2.2 Sensitive time The requirements of long sensitive time, long droplet growth time and minimising gas/aerosol diffusion losses to the walls all argue for a cloud chamber of large dimensions. Combining these considerations with the requirements of flexible expansion control and rapid turn-round time between fills, the optimum cloud chamber size is ∼50 cm linear dimension of the active volume. Before expansion, the walls and gas are in thermal equilibrium. Following an adiabatic expansion, the gas temperature is cooler than the walls, and the gas layer close to the walls begins to warm up, which re-compresses (and adiabatically warms) the inner gas volume. This sets the limit on the (static) sensitive time of the cloud chamber— the period during which no significant changes of the thermodynamic conditions occur in the central part of the chamber, where the measurements are made. For small expansion ratios, the sensitive time, 2 2 ts ∝ L /S , where L is the linear dimension of the cloud chamber and S is the supersaturation after expansion [120]. Experience with the Vienna 25 cm cloud chamber shows that ts > 10 s for S = 0.02 (2%). So the sensitive time of the 50 cm CLOUD chamber should be ts ∼3 min at 1% S, and ∼1hrat 0.2% S. However, these are static sensitive times. In the CLOUD experiment there will be a dynamical adjustment of the piston displacement to compensate for the warming effect of the walls. In this way the sensitive time will be greatly extended. Moreover, by over-compensating for the warming effect (by means of a higher rate of slow expansion) a dynamical increase of supersaturation can be generated to simulate a selected adiabatic lapse rate characteristic of a rising air parcel. These sensitive times may be compared with the droplet growth time, t ∝ r2/S, where r is the droplet radius. Measurements with the Vienna cloud chamber indicate a 1 s growth time to 1 µm radius at S = 0.02. This indicates growth times at S = 0.002 of about 10 s for droplets of 1 µm radius, and about 17 min for 10 µm radius. Therefore the sensitive time of the cloud chamber should allow measurements of droplet growth at low supersaturations up to sizes typically found in clouds. The limit is eventually −1 2 2 set by sedimentation. The terminal velocity of a water droplet at stp, vo (µms )=130 r (µm ). So, for example, a droplet of 10 µm diameter has a terminal velocity of 3 mm s−1. 5.2.3 Losses of ions, trace gases and aerosols to the walls The finite size of the cloud chamber will result in the loss of charged particles, trace gas molecules and aerosols to the chamber walls by diffusion. When an aerosol touches a wall, it attaches by van der Waals forces and is lost. For the purposes of the Monte Carlo study shown in Fig. 45 we have also assumed that a trace gas molecule or small ion is lost (or, equivalently, the ion is neutralised) if it collides with a wall. The results of the simulation indicate relatively small losses for aerosols, due to their low mobility, but significant diffusion of ions and trace gas molecules to the walls. The small ions are readily replaced by the ionising beam (§5.4), but losses of trace vapours due to wall adhesion or to aerosol growth must be compensated by a small inflow of makeup gas during long-exposure experiments. The experience of the AIDA aerosol facility [121, 122] at Forschungszentrum Karlsruhe is instruc- tive concerning the wall losses of trace gases. AIDA is a large cylindrical vessel of internal dimensions 4m(diameter) × 7m(height) with an inner ceramic lining. The temperature is controlled to 0.5 K preci- sion by cold air circulating around the outside of the vessel. It is equipped with an internal fan to ensure homogeneous mixing. Operation of the fan is found to have only a small effect on aerosol lifetimes. The measured 1/e lifetimes in AIDA for trace O3 and NO2 gases at room temperature are 70 h and 370 h, respectively. Since the lifetimes should scale as the linear dimension of the vessel (i.e. the ratio of its volume to surface area), this implies the 0.5 m CLOUD chamber should have corresponding lifetimes of about 7 h and 40 h, respectively (and four times longer for the2mCLOUDreactor chamber; §5.3.1). 50 50 10% 25% 10% 50% 25% 75% 50% 40 90% 40 75% 90% a) Small ions b) Small ions 30 293K, 101 kPa 30 223K, 26 kPa 1 min 1 min 20 20 y coordinate [cm] y coordinate [cm] 10 10 0 0 01020304050 01020304050 x coordinate [cm] x coordinate [cm] 10% 50 50 25% 25% 50% 75% 10% 50% 90% 75% 90% 40 40 c) Aerosols (10 nm diameter) d) Aerosols (10 nm diameter) 293K, 101 kPa 223K, 26 kPa 30 30 1 hour 1 hour 20 20 y coordinate [cm] y coordinate [cm] 10 10 0 0 01020304050 01020304050 x coordinate [cm] x coordinate [cm] Fig. 45: Wall losses of particles in the cloud chamber due to thermal diffusion. The upper plots show the number density of small ions after a time t = 1 minute at a) 293 K and 101 kPa (standard conditions) and b) 223 K and 26 kPa (10 km altitude). The lower plots show the number density of 10 nm-diameter aerosols after a time of 1 hour at c) 293 K and 101 kPa and d) 223 K and 26 kPa. A particle is assumed to be lost if it touches one of the walls (which are located at the boundaries of the plots). The initial charged particle distributions were generated uniformly in x and y in the range 0 5.2.4 Water vapour supersaturation The cloud chamber is required to operate over a wide range of water vapour supersaturations, S, from from below zero (unsaturated) up to about 700%. Of particular importance is the need to provide precise simulation of cloud conditions, where the supersaturation ranges from below 0.1% up to about 1%. This corresponds to a broad activation range of aerosols, namely radii from about 1 µm (at S ∼ 0.1%) down to about 50 nm (S ∼ 1%). In order to probe the CCN distribution with sufficient resolution, the cloud chamber needs to achieve a S precision after expansion of better than 0.1% in the range 0 5.2.6 Temperature and pressure range and stability The cloud chamber is required to operate over the full range of temperatures and pressures encountered by clouds in the troposphere and stratosphere, namely 185 K 5.2.7 Field cage The cloud chamber and reactor chamber are equipped with a field cage to provide an electric field within the active volumes. The field cage is important for several reasons. In its simplest application, a modest electric field of about 1 kVm−1 will clear small ions from the cloud chamber in about 2 s. This is well below typical charge attachment times onto aerosols and will allow control measurements to be made at effectively zero ionisation (below 1% of ground-level ionisation). A second application is to select the sign of the ions or charged aerosols before droplet activation, in order to allow investigation of sign- dependent effects. This can be achieved by setting the electrode voltages to produce an electric potential ‘valley’ for the selected sign at the mid-point of the chamber (and a ‘ridge’ for the opposite sign). The practical field will more resemble an inverted saddle (or saddle) and so there will still be some loss of the selected sign towards the walls at the mid-point of the chamber. However, a strong sign-selection can be achieved. A third application of the field cage is to establish realistic electric fields for the cloud electricity experiments (§4.6). The latter set the desired maximum electric field to about 10 kVm−1, which covers all cloud conditions except lightning (Table 3). Finally the field cage may provide useful information on the magnitude of the droplet charge in situ.This involves a Millikan-type measurement of the terminal velocity of a droplet vs. electric field. It is limited to droplets that are visible by the CCD cameras (i.e. above about 1 µm diameter) and have sufficient charge to be suspended, or almost suspended. For example, in an electric field of 10 kVm−1, 3 3 the charge required to suspend a droplet is qs(e)=24r (µm ), where r is the droplet radius. This indicates minimum charges of 3e are required to suspend a 1 µm diameter particle and 24e for a 2 µm diameter particle. In dipolar small-ion clouds, the typical droplet charges from diffusion will be a few electronic charges, so this technique appears to be useful only for particles in a narrow diameter range of afew-µm’s. It may also be possible to devise useful ways to estimate aerosol mobility (charge/size) in situ by drifting the aerosols for a known time before droplet activation. The electric field is provided by a field cage attached to the walls of the cloud and reactor chambers. A similar design has been successfully used in the large TPC tracking detector of ALEPH at LEP. The ALEPH TPC involves a (half) cylinder of 2.2 m length and 3.6 m diameter, with an electric field of 11.5 kVm−1 that is provided by field cage electrodes on the surface of the cylinder, with none placed in the active (half) volume. 5.2.8 Chamber cleaning CLOUD experiments will frequently involve extremely low concentrations of trace vapours (around 1 pptv, equivalent to concentrations of 107 molecules cm−3 or less) and aerosols (1cm−3). Therefore trace impurities must be eliminated at well below this level. Fortunately, provided suitable care is taken, the cloud chamber is self cleaning [119]. By repeated and progressively deeper expansions, impurities are activated and can be sedimented out of the active volume (the same can be achieved in the reactor chamber by vacuum pumping). Impurity concentrations of <0.01 condensation nuclei cm−3 can be reached, where a condensation nucleus refers to a molecular cluster of any size. However, to reach this performance, the chamber must be carefully cleaned between each experiment. In most cases the cleaning of aerosols and gas molecules attached to the walls can be carried out by vacuum baking. This involves first emptying the liquid fluorocarbon coolant and then vacuum pumping of the active volume while heating the chamber by means of heater cables wrapped around the inner chamber body. Periodically, it will be necessary to flush and clean the inside of the chamber with liquid solvents. This can be achieved in two ways. The first makes use of a central hole through the piston drive rod and piston. This can be used both to flush liquids into the chamber and also to drain them. The second involves a more thorough cleaning (or repair) by disassembly of the cloud chamber body. The latter is designed to be able to be warmed up, disassembled, repaired, reassembled and cooled down again in 24 hours. During such an operation, the lower cloud chamber assembly, including the piston and actuator system, is left intact. scintillation counter roof (GCR monitor) cooling /temperature control gas & aerosol systems cryogenics liquid FC synthetic air/ argon cooling UV illumination water vapour mixing field cage voltage chamber aerosols in situ analysers inspection video camera trace gases CCD cameras Mie scattering detector ice particle lasers detector beam retractable telescope aerosol /trace probe gas analysers refractive index gas thermometer temperature condensation particle & pressure counters (CPC) differential mobility particle sizers (DMPS) trace gas analysers mass piston control piston vacuum spectrometers system actuator system ion mobility expansion spectrometers system external analysers Fig. 46: Schematic diagram of the subsystems and instrumentation for CLOUD. Normally the gas/aerosol samples for the external analysers are drawn from the reactor chamber rather than the cloud chamber as indicated the figure. 5.3 Experimental design 5.3.1 Cloud chamber and reactor chamber The CLOUD detector is shown in Figs. 43 and 44, and the subsystems and instrumentation are shown schematically in Fig. 46. The main components are a 0.5 m cloud chamber anda2mreactor chamber. The purpose of the reactor chamber is twofold: i) it provides the samples of reacted gas/aerosols for analysis by the external detectors, and ii) it is required for long growth-time experiments lasting several days, since it has about a factor four longer particle lifetimes than the cloud chamber, due to reduced wall losses. For these experiments the reactor chamber can also provide samples of reacted gas/aerosols for analysis in the cloud chamber since it has a much larger volume (by a factor 64). The chambers are exposed to a muon beam of large transverse dimensions (about 2 × 2 m2) which provides a simultaneous equal ionisation of both volumes. The chambers are constructed from aluminium, with black teflon lining the inner surfaces. The teflon is made partly conductive in order to prevent any charge accumulation. Both chambers can be operated at pressures between a vacuum and 1.5 atm. Precision pressure gauges and Pt thermometers are set into the chamber walls. A teflon-coated field cage is suspended inside each chamber, isolated from the walls by standoff insulators. The field cages generate vertical electric fields inside the chambers of up to 10 kVm−1 with a flexible choice of field profiles. The piston expansion system for the cloud chamber comprises the main piston, drive rod and actuator. Two options are under study for the actuator system. One option is a hydraulic system using aservo valve, as shown in Figs. 43 and 44 and described in the CLOUD proposal [3]. It is based on the same design as was used to control the 2 m-diameter piston of BEBC [124]. The second option is a linear electric servo motor with rare-earth permanent magnet arrays attached to the drive rod and liquid- cooled iron-core electromagnetic coils fixed to the support structure [123]. Both options provide very precise displacement of the piston, and flexible electronic control of any desired expansion/compression cycle via digital signal processors. The most rapid expansions can be made under 200 ms for a 35% volume expansion. Precision sensors measuring temperature, pressure and piston displacement provide feedback to control the piston actuator. The piston is a stiff and lightweight sandwich assembly. A critical component is the piston seal. The present design foresees a Bellofram rolling diaphragm [125], with a permanent vacuum below the piston [123]. A bias piston at the base of the drive rod coupled to a large pressurised ballast tank compensates most of the excess downward force on the piston (2.0 tonnes at 1 atm). A small depression in the top surface of the piston allows for a thin pool of water (or ice) to establish 100% relative humidity. The liquid cooling and temperature control system involves a closed circuit system, insulated throughout by vacuum. The liquid fluorocarbon coolant flows through a jacket surrounding the cloud chamber and reactor chamber and maintains the inner walls and piston at a precisely-controlled temper- ature. The gas inside the cloud chamber is allowed to reach thermal equilibrium with the walls before taking any measurements. Heater cables are also wrapped around the chamber vessels to allow cleaning by vacuum bakeout. The reactor chamber is operated at the same temperature and pressure as the cloud chamber. However, since no droplet activation is involved, a relatively modest temperature stability of a few × 0.1 K is adequate. The gas and aerosol supply systems involve four components: carrier gas, water vapour, aerosols and trace gases. The carrier gas is either pure artificial air (80% N2, 20% O2)orargon. Water vapour, aerosols and trace gases are mixed into this stream at the desired levels . The water vapour content in the chamber will be set by two techniques: either a) a liquid or ice film on the top of the piston or b) vapour introduced from an external humidifier. Aerosol particles will be generated with standard techniques such as nebulizers. Care must be taken to minimise transmission losses between the aerosol generators and the chamber volumes (and between the sampling probes and the external analysers). In general, sub-micron particles are quite efficiently transported in small tubes at standard carrier-gas flow rates. The main mechanisms for transport losses are diffusion onto the tube walls, gravitational settling, and inertial losses at sharp bends. Charged particles have additional losses due to electrostatic attraction. Diffusion losses are only significant for particles smaller than approximately 20 nm. Gravitational losses are usually only significant for particles exceeding about 0.5 µm diameter. All these losses, and probe sampling efficiencies, are well known [126] and can be accounted for. Both chambers are equipped with gas/aerosol inlet and outlet pipes, as well as ports connecting to the vacuum system and connecting each chamber to the other. The inner vessels are fitted with sampling probes to extract gas and aerosols for the external analysers. The tip of each sampling probe can be inde- pendently adjusted to any radial position inside the vessel (these are visible inside the reactor chamber in Fig. 44). A small fan is installed inside the reactor vessel to provide the option of slow stirring of the gas filling to assist, where necessary, homogeneous mixing throughout the large volume. A special inlet pipe located near the fan provides fresh trace gas to replace losses to the walls or to aerosol growth. As well as windows for optical readout, the chambers are equipped with inspection and illumination windows. Internal UV lamps are also provided for experiments involving gas-phase photochemistry. 5.3.2 Instrumentation The CLOUD instrumentation is shown schematically in Fig. 46. The in situ analysis of the cloud chamber comprises the following systems: a) a constant angle Mie scattering (CAMS) detector [127], b) a stereo pair of CCD cameras, c) an ice particle detector which measures backscattering of a polarised laser beam [122], d) gas and aerosol analysers, e) a refractive index gas thermometer based on a laser interferometer [128], and f) precision temperature and pressure monitors set into the inner wall. During activation, the cloud droplets are primarily analysed by the CAMS and CCD systems. These are complementary but, nevertheless, have a broad region of overlap where they can provide mu- tual cross-checks. The CAMS system can measure very high droplet number densities (∼10–107 cm−3) whereas the CCD cameras operate best in a lower range (∼0.1–105 cm−3). The CAMS system provides a high-resolution measurement of mean droplet radii vs. time, whereas the CCD cameras provide a mea- surement of droplet size in coarser time intervals, using pulse height information. Finally, the CAMS detector integrates over all illuminated droplets whereas the CCD cameras reconstruct the 3-dimensional spatial positions of individual droplets, and track their movements. This is important for identifying ice nuclei and for measuring droplet drift and sedimentation trajectories. The illumination system for the CAMS and CCD systems comprises: a) a laser for illumination of a narrow region for the CAMS detector (and, in parallel, for the CCD cameras) and b) a xenon flash tube mounted at the top window, for the CCD cameras. A video camera is also mounted at the top window to provide a visual inspection of the chamber volume and piston surface. All windows are made of optical quality quartz (since it is UV-transparent and has a high thermal conductivity, similar to stainless steel). As well as in situ analysis, samples of gas are drawn from the chambers via retractable probes and directed to an array of external instruments to analyse the chemical and physical characteristics of the aerosols, trace gases and ions. The instruments include condensation particle counters, differential mobility particle sizers, trace gas analysers, mass spectrometers and ion mobility spectrometers. The beam intensity and profile is measured by a plastic scintillation counter telescope. The counters cover the full 2 m width of the beam and include a fine-grained array of 8 × 8 counters of size 25 × 25 cm2 to measure the transverse profile. Finally, a roof of plastic scintillation counters monitors the GCR exposure, which will be significant for measurements taken at the lowest beam exposures. More details of the instrumentation for CLOUD can be found in ref. [3]. 5.4 Particle beam 5.4.1 Why a particle beam? The basic experimental requirement of CLOUD is to duplicate atmospheric and cosmic ray conditions in the laboratory. Essentially throughout the troposphere, charged cosmic rays are minimum ionising par- ticles (§4.3). These are mostly relativistic protons in the upper troposphere (above about 7 km altitude), and muons in the lower troposphere (Fig. 26). The requirements of the ionisation source for CLOUD are as follows: • Deposition of a precisely known quantity of ionisation within the cloud and reactor chambers. • Uniform ionisation over a large volume (about 2 × 2 × 2 m3). • An ionisation density (dE/dx) that is characteristic of minimum ionising particles. • Easily adjustable in intensity over the required range of 1–10× the natural cosmic ray intensities found in the troposphere—an intensity range of about 1000. • Ability to traverse the walls and liquid cooling layers of the cloud chamber and reactor cham- ber. This sets a minimum energy for a particle beam of about 1 GeV/c, taking multiple Coulomb scattering also into account. • Known timing. This is necessary for the study of fast processes and also for ice nucleation studies to distinguish between deposition nucleation and freezing nucleation. Fig. 47: An image of X ray interactions in a cloud chamber recorded by C.T.R. Wilson in 1912 [74]. The beam travels horizontally and has a diameter of about 2mm. The horizontal field of view is about 14 mm. Corrected to one atmosphere pressure, Wilson counted about 200 ion pairs per cm for the less highly ionising regions of these tracks, and over 2000 ion pairs per cm in the highly ionising regions near the ends of the tracks. Fig. 48: An image of minimum ionising particles recorded by C.T.R. Wilson in 1912 with the same cloud chamber and operating conditions as Fig. 47 [74]. Two crossing tracks are seen, indicated by the arrows. The horizontal field of view is about 18 mm (the same scale as Fig. 47). Corrected to one atmosphere pressure, Wilson counted about 32 ion pairs per cm for these tracks, including the δ rays (the tightly packs groups appearing as bright spots). Other sources of ionising radiation include ultraviolet (UV) radiation, radioactive sources and Xray sources. UV radiation is excluded since it induces photochemical reactions among the trace gases. In the case of radioactive sources, α emitters are excluded by their high ionisation density and short range, and β emitters are excluded since they cannot be placed inside the cloud chamber, and their range is insufficient to penetrate the chamber walls. Gamma radioactive sources and energetic X rays can penetrate the chamber walls but their ionisation deposition is dominated by stopping electrons, which have much higher dE/dx than minimum ionising particles. This can be seen visually by comparing Figs. 47 and 48, which show some remarkable cloud chamber images of X rays and minimum ionising particles, respectively, recorded by C.T.R. Wilson in 1912. In summary, a particle accelerator beam is the only source of ionising radiation capable of realistically duplicating cosmic rays in the laboratory. In fact some little-known laboratory studies of ion-induced effects on aerosol formation have been carried out since the 1960’s using traditional ionisation sources (e.g. refs. [97, 98, 129]). Members of the CLOUD collaboration have also studied these processes more recently using X rays and α particles from 241Am sources [130]. Although some useful results have been obtained, these studies have generally been unable to characterise the aerosol processes adequately. The main limitations have been a) lack of control of the ionisation and ionisation rate (dE/dx)atnear-atmospheric intensities, and b) non-uniformities of deposited ionisation. It has been shown [131] that local variations of ion density (such as from an α source) give rise to non-linear aerosol charging effects, which will directly affect ion-induced aerosol processes. This makes it difficult to relate results obtained with such sources to the real atmosphere. Table 4: Summary of the approximate minimum and maximum time-averaged beam intensities for CLOUD. The GCR intensity at ground level is about 0.02 cm−ss−1, which is included in column 3 of the table. A transverse 2 beam size of 200 × 200 cm is assumed. IGCR signifies the natural GCR intensity at the indicated altitude. Beam intensity Beam + GCR intensity Clearing Simulated GCR conditions (s−1)(cm−2s−1) (cm−2s−1) field intensity altitude (km) 00 0.02 on 0.01 × IGCR 0 00 0.02 off 1 × IGCR 0 3 10 0.02 0.04 ” 2 × IGCR 0 4 10 0.2 0.22 ” 10 × IGCR 0 00 0.02 ” 0.01 × IGCR 20 5 10 2 2 ” 1 × IGCR 20 6 10 20 20 ” 10 × IGCR 20 5.4.2 Beam requirements The beam requirements for CLOUD are as follows: • A muon beam; the ideal particles are µ’s since they replicate the GCR’s in the lower troposphere, and do not interact in the detector material, as would π’s or p’s (e’s are excluded due to showering). • A mean energy of several GeV, to avoid large multiple Coulomb scattering. The energy spread should be known (to calculate the ionisation energy deposition, dE/dx), but it can be broad. • An adjustable, time-averaged intensity from about 103 s−1 to 106 s−1 (and also zero beam). • Precise (few per cent) knowledge of the beam intensity (since the solar modulation corresponds to arather small change of GCR intensity of only ∼10%). • A large transverse size, 200 × 200 cm2. • Continual operation throughout the year. A typical single experimental study will require about 4 weeks beam-time, and many experiments are foreseen (§5.5). The desired beam intensity is between about 1× and 10× the natural GCR flux at any altitude. This will allow measurements of the dependence of any observed effects on ionisation rate and ion pair concentration at the natural ionisation rates in the atmosphere. The highest beam intensities will help to amplify and expose effects before they are measured at natural ionisation levels. Beam-off data will be also be recorded under conditions with the chamber clearing fields on and off, respectively, corresponding to 0.01× and 1× the natural GCR ion-pair concentrations at ground level. CLOUD will measure processes over the full range of tropospheric and stratospheric conditions. At ground level, the average GCR intensity is about 0.02 cm−2s−1, whereas at 15–20 km altitude it is about a factor 100 larger, varying between about 0.8 and 2.3 cm−2s−1 depending on geomagnetic latitude (Fig. 27a). The maximum required time-averaged beam intensity is therefore about 10 × 2= 20 cm−2s−1. The beam is spread over a large transverse area of 200 × 200 cm2 in order to duplicate the quasi-uniform GCR irradiation, over the fiducial volume. The time-averaged maximum beam intensity is then 20 × 4 · 104 106 s−1. The minimum beam intensity (apart from beam-off) is about 1× the natural GCR radiation at ground level. This is a factor 1000 below the maximum required intensity (a factor 100 for the atmospheric attenuation and a factor 10 for 1× the GCR intensity rather than 10×), i.e. a time-averaged intensity of 103 s−1. These beam estimates are summarised in Table 4. 5.5 Experimental programme overview 5.5.1 Aerosol experiments Gas-to-particle conversion: The aim is to measure the effect of ionising particle radiation on the formation rate of ultrafine condensation nuclei (UCN) in the few-nm size range from trace precursor vapours. Such trace gases include, in particular, H2SO4, HNO3,NH3 and certain volatile organic com- pounds, all in the presence of H2Ovapour. The basic parameter to be measured is the UCN nucleation rate, J (cm−3s−1), as a function of the primary experimental variables: trace vapour concentration, rel- ative humidity, temperature, background aerosol concentration and ion-pair production rate. The cloud chamber can measure J over a wide range from about 3 · 10−5 cm−3s−1 to 107 cm−3s−1.Theoretical studies [84, 85] indicate that large differences are expected in the ion-induced nucleation rate between positive and negative charges, by factors of up to 100 or more. These effects will be studied either with a ‘saddle’ electric field. Typical formation times (i.e. beam exposures) for these experiments of up to about one hour are expected. Growth of CN into CCN: The purpose of these experiments is to measure the effect of ionising particle radiation on the rate of growth of CN into CCN (i.e. from ∼5nmdiameter to ∼100 nm) in the presence of condensable vapours that are known to be important in the atmosphere, such as sulphuric acid, water, ammonia and organic compounds. Both ‘dry’ and aqueous-phase growth will be studied. The latter will involve expansion and compression cycles of the cloud chamber to activate droplets and then evaporate them. The basic parameter to be measured is the CCN concentration, CCN(S).Aprogressive expansion will be used to provide a gradually-increasing supersaturation so that the CCN concentration can be measured as a function of S. The CCN are the subset of the CN population that activate into droplets at agiven water vapour supersaturation, S. The fraction of the CN that constitute CCN depends not only on their size but also on their chemical composition and other properties. The evolution of the CN size spectra during beam exposure will also be measured. A typical single run will last up to about 2 days. Activation of CCN into cloud droplets: The goals of these experiments are to see how the presence of electrical charge affects the critical supersaturations required to activate CCN, and also how it affects the number of cloud droplets that appear. Well-defined aerosol sizes will be produced with standard aerosol generation techniques. Experiments will be performed to examine the activation of these aerosols into cloud droplets. The charge distribution on the aerosols will be measured in situ using the field cage, and externally with the ion mobility spectrometer. A wide range of CCN and ambient conditions (background aerosol, trace gases and thermodynamic conditions) will be explored. The simplest systems will involve typical hygroscopic aerosols found in the atmosphere, e.g. NaCl, (NH4)2SO4 and H2SO4, and also partially soluble aerosols such as CaSO4. Other measurements will be made on hydrophobic carbonaceous particles, with and without the presence of organic surfactants. The aims here are to investigate the effect of ionisation on the reduction of the critical supersaturations and on the increases of droplet number associated with the effects of surface charges [132] and surface tension suppression [54]. Finally, these measurements will be repeated in the presence of additional, highly-soluble, trace gases, which are expected to have an extremely low critical supersaturation, even to the point of activating into cloud droplets at below 100% relative humidity [133]. Dynamics of CCN activation and kinetic limitations on cloud droplets: The purpose of these ex- periments is to study the number of droplets that activate from a given CCN population under various dynamical supersaturation conditions. The calculation of the number of cloud droplets that will activate from a given CCN concentration is generally calculated under the assumption the classical Kohler¨ theory (§4.4.1). However this assumes an equilibrium situation whereas, in real clouds, aerosol growth before activation may be kinetically limited [134]. The result can be large errors in the calculated number of cloud droplets. Since the radiative forcing of clouds is very sensitive to droplet number (§4.4.2), it is im- portant to be able to determine the concentration of cloud droplets to a few per cent. These experiments will first involve the preparation of a known, polydisperse, CCN population, with various CCN concen- trations and compositions. The number of droplets that activate from these different CCN populations will then be investigated in the cloud chamber under different realistic adiabatic lapse rates, simulated by the appropriate progressive expansion cycles, both with and without beam ionisation. Production of NO and OH and their influence on aerosols: These experiments will first quantify the poorly-known production rates of a) nitric oxide (NO) and b) hydroxyl radicals (OH) by cosmic rays. Both are important reactive chemicals in the atmosphere and, in certain regions, production by GCRs may be a significant, or perhaps the dominant, mechanism. Estimates made over twenty years ago [68, 69] suggest about 1–2 OH radicals and 1.5 NO molecules are produced per ion pair. These production rates would imply maximum mixing ratios of around 1 pptv per day. The production of NO will be measured in pure artificial air and the production of OH will be measured in argon and water vapour. In a second step, the effect of GCR-produced NO and OH on aerosol nucleation, growth and activation will be studied by comparing measurements taken with two different carrier gases: artificial air and argon. Production of aerosol precursors from trace gases: The aim of these experiments is to quantify the influence of ionising radiation on various rate-limited chemical reactions that are important for atmo- spheric production of aerosol precursors. For example, sulphuric acid vapour is among the most impor- tant aerosol precursor gases for CCN production. It is mainly produced by oxidation of SO2, which in turn may be emitted directly or else as a product of DMS oxidation (Fig. 35). However the oxidation of SO2 in the atmosphere is slow since it proceeds by interaction with the rare hydroxyl radical. The lifetime 5 6 −3 of SO2 by the OH reaction is about 1 week at typical OH concentrations (10 − 10 molecules cm ). In contrast, the lifetime of SO2 by dry and wet deposition is about a day. Any process that apprecia- bly increases the rate of SO2 oxidation will therefore increase the H2SO4 concentration. Since GCRs produce OH radicals, they can in principle influence this reaction. There is in fact some experimental ev- idence that ionising particles may indeed enhance the rate of H2SO4 formation [82]. These experiments will involve exposing various atmospheric concentrations of trace gases under different thermodynamic conditions to a range of beam fluxes, and then analysing the reaction products. Effect of highly ionising particles: Although the majority of cosmic rays in the lower troposphere (below about 6 km) are muons, a few per cent are protons. These undergo nuclear reactions, generating highly ionising nuclear fragments. The fraction of strongly interacting particles (p, n, π± and heavy ions) increases at higher altitudes. It is therefore important to investigate the effect of highly ionising fragments on the nucleation of aerosols (and perhaps on other processes summarised here) [135]. These experiments will involve the introduction of small amounts of radon gas into the chamber gas mixtures. Radon isotopes are α emitters, e.g. 222Rn decays to 220Po and a 5.5 MeV α, with a 3.8 d half life. The α particle range in air is about 5 cm at stp, corresponding to an ionisation density, dE/dx, that is about afactor 500 larger than it is for muons (minimum ionising particles). The data taken with Rn irradiation will be compared with beam data to study the effects of high ionisation densities. 5.5.2 Ice particle experiments Ice particle formation by deposition nucleation: The aim is to study the influence of ionisation on the ability of aerosols to act as ice nuclei (IN) on which water vapour can be directly sublimated to the solid phase (Fig. 38a). The cloud chamber will be filled with selected aerosols and exposed to the beam. The beam is then turned off and expansions made to a range of final temperatures between 0◦C and -40◦C. The efficiency of these aerosols to act directly as IN will be measured as a function of temperature and aerosol charge (using the electric field created by the field cage). The presence of an ice crystal in a cloud of drops is readily identified by the CCD cameras since an ice crystal scatters much more light than a water drop [103]. The ice fraction will also be measured with a backscattered polarised laser beam. Ice particle formation by freezing nucleation and contact nucleation: These experiments will in- vestigate the influence of ionising radiation on the formation of ice particles via the intermediate stage of supercooled liquid droplets. The expansion chamber will be used to create a supercooled cloud by ex- pansion and growth of droplets at temperatures between 0◦C and -40◦C. The temperature of the droplets can be controlled by the expansion ratio and the initial temperature of the chamber before expansion. After forming the droplets, they are exposed to the beam and the freezing rate is measured. There are several ways to disentangle freezing nucleation (Fig. 38b) from contact nucleation (Fig. 38c). The time development will be different for the two processes (contact nucleation is slower since it is controlled by diffusion of low-mobility aerosols and by sedimentation) and, in addition, CCN that do not act as IN can be included to provide the condensation nucleus for the supercooled droplets. It is important to observe the freezing process over an extended time. This can be achieved by dynamically adjusting the expansion piston so that the supercooled droplets neither evaporate nor grow so large that they are lost by sedimentation. Secondary production of ice particles: These experiments will study the effect of charge on the pro- duction of secondary ice particles by shattering and splintering of freezing droplets. Mixed-phase clouds will be created at various temperatures and the particles grown to sizes where sedimentation and colli- sions take place. The secondary production of ice particles by contact freezing will be studied. These measurements will use the CCD cameras to follow the evolution of individual particles and their interac- tions with other particles. Efficiency of highly-charged evaporation nuclei as IN: These experiments will study the Tinsley mechanism (§4.5.2). This will first involve forming cloud droplets at an appropriate temperature in the range from 0◦Cto-40◦C, in the presence of selected initial aerosols and trace vapours, including both sol- uble and insoluble (organic) compounds. The droplets will be charged to relatively high values, ∼ 100e, using the beam and electric field. The droplets will then be evaporated by an adiabatic compression, and the cycle repeated. In this way the effectiveness as IN will be investigated of highly charged aerosols coated with cloud-processed material. Effect of freezing-thawing cycles on IN efficiency: These experiments will study the effect of freezing- thawing cycles on the efficiency with which various aerosols act as IN. After filling the cloud chamber with the aerosols under study, a series of expansion/compression cycles will be used to activate droplets, allow some to form ice particles, and then evaporate them and repeat the cycle. Growth of ice particles in mixed-phase clouds: The aims of these experiments are to measure the dynamics and ice-particle growth rates in clouds formed of co-existing liquid and ice particles. Clouds at various temperatures between 0◦C and -40◦C will be generated containing liquid particles and a few ice particles. The latter will be created using efficient IN such as AgI, in the appropriate low ratios relative to the CCN (10−4 − 10−6). Reflectivities of ice and liquid clouds: The aim here is to measure optical reflectivities (albedo) of ice clouds and liquid clouds as functions of water content and particle number concentrations. Freezing mechanism of polar stratospheric clouds: The aim of these experiments is to investigate the deposition freezing nucleation of HNO3 and water vapours onto ion clusters, forming nitric acid hydrates. Particles composed of such hydrates are the principal component of the polar stratospheric clouds that initiate the destruction of ozone. The operating temperatures are typical polar stratospheric values of between 190 and 200 K. Nitric acid and water vapour will be present the chamber at partial −4 −2 pressures representative of the stratosphere (10 Pa for HNO3 vapour and 5 · 10 Pa for H2Ovapour, corresponding to about 10 ppb and 5 ppm, respectively). At these pressures and temperatures the nitric acid hydrates become supersaturated and can condense as crystals provided a suitable ice nucleus is present. Sulphuric acid vapour will be included in the air mixture to represent the species most likely to contribute to the initial ion cluster formation. 5.5.3 Cloud electricity experiments Effect of charge and electric field on the coalescence efficiency of droplets: The aim of these ex- periments is to study the effect of charge and electric field on the efficiency with which cloud droplets coalesce on colliding. Laboratory studies [136] have indicated that raindrops (of around 0.5 mm diame- ter) are about a factor 100 more efficient at collecting aerosol when they are charged rather than neutral, and theoretical studies support this picture [105]. Since collisional growth is the main mechanism by which cloud droplets grow into raindrops, this is an important process for determining the lifetime of a cloud. These experiments will involve generation of polydisperse droplet distribution in the cloud cham- ber (from a broad initial CCN distribution). The coalescence efficiency will be measured for individual droplets using the CCD camera system, as a function of droplet size, charge and electric field. In addi- tion, special runs will be taken with high ionisation depositions to simulate the ion densities that may occur in thunderstorm conditions, where the coalescence efficiencies may be greatly enhanced. Charge separation mechanism in clouds: The aim of these experiments is to investigate whether there are any differences in critical supersaturation between positively and negatively charged CCN. Theoretical studies [112] suggest that negative particles may activate more readily than positive particles. If this occurs under natural cloud conditions then it could lead to charge separation, the precursor for lightning. In these experiments the beam will first ionise the cloud chamber volume and the charge allowed to be scavenged by a predetermined CCN population. The beam will be turned off and the clearing field used to separate the positive and negative charged aerosols. A progressive piston expansion will gradually raise the supersaturation. The activation of the positively-charged CCN region can then be compared with the negatively-charged CCN region. After activation, the charges on the aerosols can be measured by further drifting in an electric field set with the field cage. The activations will be measured as functions of the CCN type, beam intensity, aerosol charge, electric field and adiabatic lapse rate (rate of rise of supersaturation). Lightning trigger mechanism: The aim here is to investigate the proposed triggering mechanism for lightning due to high energy primary cosmic rays [113]. The charge amplification in the cores of these showers will be studied under typical cloud conditions (i.e. cloud droplets and electric fields). A single pulse of about 2 · 104 beam particles over a 2 × 2 m2 area simulates the peak electron shower density in the core of 1015 eV primary GCR. Higher particle densities can easily be generated. The field cage generates electric fields of up to 10 kVm−1,which will simulate the fields in all but lightning clouds. Higher fields typical of lightning breakdown conditions (100 kVm−1)can be generated using specially designed electrodes inserted into the chamber via the sampling probes. For these experiments an X ray detector will be used to observe possible X ray emissions from avalanche processes. Effect of electrical discharge on cloud droplet formation: The aim of these experiments is to study the effects of electrical discharges on cloud droplet number and on droplet coalescence efficiency, and especially to quantify the effects of NOx production from electrical discharges. Since GCRs may influ- ence the rate of lightning, they may in turn influence this important source of NOx in the atmosphere. It is interesting to note that cloud chamber experiments in Edinburgh in the 1960’s observed backgrounds due to ‘hypersensitive condensation nuclei’ [137]. These activated into droplets at very low water vapour supersaturations—of order 1%—and were attributed to trace amounts of NO2 vapour created by electro- static discharges. As before, the discharges will be generated by specially designed electrodes inserted into the cloud chamber via the sampling probes. 5.5.4 Data interpretation and cloud modelling The experimental results obtained with CLOUD will be evaluated with aerosol and cloud simulations. The basic physics aims of these simulations are as follows: 1. To incorporate the microphysics of ion-mediated processes into the aerosol and cloud models. 2. To examine the sensitivity of clouds under atmospheric conditions to variations in the GCR inten- sity, in the presence of other sources of natural variability. There will be a close feedback between the simulations and the experimental observations to confirm a) that the simulations closely reproduce the experimental data and b) that the underlying micro- physics is understood. The simulations will also eventually become important in guiding the direction of the experimental programme. As well as a detector simulation, this work will require a complete simulation of the microphysical processes under study. Already existing within the CLOUD collaboration are detailed models of aerosol nucleation, growth and activation, and also two sophisticated cloud models—for cumulus and marine stratus clouds, respectively. Ion-mediated microphysics will be incorporated into these models based on the experimental data from CLOUD. If item 2 above reveals discernible GCR effects on single clouds, suitable parametrisations will be prepared for the climate modelling community to explore the influence of GCR effects on the global climate. 6 CONCLUSIONS The Sun is a variable star. The sunspot record over the last 400 years reveals both a quasi-periodic solar cycle of about 11 years and also longer-term changes, including a grand minimum lasting about 70 years during which the sunspots all but disappeared. This coincided with the most pronounced of several prolonged cold spells between 1450 and 1890 which are collectively known as the Little Ice Age. Light radio-isotope records in ice cores, tree rings and other archives extend the measurements of solar variability back over the last 250 kyr. They reveal numerous occasions when the Sun waxed or waned over centennial periods. The Earth’s climate is far from stable. Natural forcings have caused large climate changes in the past. The most important are the periodic shifts from glacial epochs lasting about 100 kyr to warm inter-glacial epochs lasting about 10 kyr. These transitions involve global average temperature changes of about 5◦C, and polar changes of 10–15◦C. They coincide with the Milankovitch orbital variations of the Earth around the Sun, but amplification mechanisms are needed to account for the magnitude of the observed climate changes. However, in addition to these large climate shifts, recent palaeoclimatic studies have shown that significant climate variations also occurred within the Holocene and the last ice age on centennial and millennial timescales. There is increasing evidence that many of these are triggered by solar forcing. It appears that the Little Ice Age was but one of around 10 solar-induced cold spells during the last 10,000 years. However, despite the extensive evidence, solar variability remains a controversial agent of climate change since no causal mechanism has been established to link the two phenomena. Estimates of the variation of the solar irradiance appear to be too small to account for the observed climate variability. Until recently there were no indications where else to look. However, recent satellite data have provided anew clue: low cloud cover may be influenced by the Sun—either as a result of changes in the elec- tromagnetic (UV) radiation or in the galactic cosmic rays, whose flux is modulated by the solar wind. This could provide an effective initial step by which an energetically-weak solar signal is amplified into a significant climate forcing. Subsequently, in response to the cloud radiative and hydrological changes, further amplification processes may occur such as shifts of the global thermohaline system, which can exert major global climatic effects. The cosmic ray hypothesis for solar-climate variability rests on the microphysical interactions of cosmic rays and clouds, which are experimentally poorly-known. The CLOUD experiment proposes a major investigation of ion-aerosol-cloud microphysics under controlled laboratory conditions using a beam from a particle accelerator, which provides a precisely adjustable and measurable artificial source of cosmic rays. The heart of the experiment is a precision cloud chamber that recreates cloud conditions throughout the atmosphere. CLOUD is designed as a flexible, general purpose detector, and a broad range of measurements of cloud microphysics is planned in the areas of aerosols, ice particles and atmospheric electricity. Unique in the world, this facility would open up an essentially new line of atmospheric research. Its primary task is to pursue the question of how cosmic rays may influence clouds. If clouds respond to the solar variations that modulate the cosmic rays reaching the Earth, there are consequences for the evaluation of climate change, since the 20th century warming coincided with a large increase of solar activity. To settle the issue, one way or the other, is therefore of great importance. Finally, since this paper begins with a quote, it is perhaps fitting to end with another [138]: The physics of cloud formation appears in every aspect of the climate variation, includ- ing direct heating by sunlight. I would hope that with the important issues that need to be decided in connection with climate variation and global warming, an all out effort can be launched to understand this vital aspect of atmospheric science. There is a lot of laboratory work to be done under carefully controlled conditions, e.g. the CERN [CLOUD] experi- ment, on the nucleation of aerosols, ice crystals, and water drops, as well as the related field work to study the application of the laboratory results. ... Our ability to handle the scientific challenge of climate change will be a subject of future historical research and writing. The eyes of future generations will be upon us. There has never been an individual scientific problem that will have so much impact as the global warming inquiry on the long term well being of the human race. Eugene N. Parker, University of Chicago Conference Summary, The Solar Cycle and Terrestrial Climate Santa Cruz de Tenerife, September 2000 Acknowledgements It is a pleasure to thank Jurg¨ Beer, Gordon Bowden, Nigel Calder, Maurice Jacob, Lew Keller, Ralph Nel- son, David Ritson, Arnold Wolfendale and my CLOUD collaborators for many interesting and enjoyable discussions. References [1] H. Svensmark and E. Friis-Christensen, Variation in cosmic ray flux and global cloud coverage—a missing link in solar-climate relationships, Journal of Atmospheric and Solar-Terrestrial Physics 59 (1997) 1225. [2] N. Marsh and H. Svensmark, Low cloud properties influenced by solar activity, http://arXiv.org/abs/physics/000507, Phys. Rev. Lett. 85 (2000) 5004–5007. [3] CLOUD collaboration (B. Fastrup et al.), A study of the link between cosmic rays and clouds with a cloud chamber at the CERN PS, http://www.cern.ch/cloud/iaci workshop/cloud.html, CERN proposal SPSC/P317, SPSC 2000-021 (2000). [4] CLOUD collaboration (B. Fastrup et al.), Addendum to the CLOUD proposal, http://www.cern.ch/cloud/iaci workshop/cloud.html, CERN SPSC/P317 Add.1, SPSC 2000- 030 (2000). [5] CLOUD collaboration (B. Fastrup et al.), CLOUD: an Atmospheric Research Facility at CERN, http://www.cern.ch/cloud/iaci workshop/cloud.html, CERN SPSC/P317 Add.2, SPSC 2000-041 (2000). [6] J.M. Wilcox, Solar activity and the weather, J. Atm. & Sol.-Terr. Phys. 37 (1975) 237–256. [7] W. Herschel, Observations tending to investigate the nature of the Sun, in order to find out the causes or symptoms of its variable emission of light and heat, Philosophical Transactions of the Royal Society,London (1801) 265. [8] J.A. Eddy, The Maunder minimum, Science 192 (1976) 1189. [9] C. Frohlich¨ and J. Lean, The Sun’s total irradiance: Cycles and trends in the past two decades and associated climate change uncertainties, Geophys. Res. Lett. 25 (1998) 4377–4380. [10] J. Lean, J. Beer and R.S. Bradley, Reconstruction of solar irradiance since 1610: Implications for climate change, Geophys. Res. Lett. 22 (1995) 3195–3198. [11] M. Lockwood, R. Stamper and M.N. Wild, A doubling of the Sun’s coronal magnetic field during the past 100 years, Nature 399 (1999) 437. [12] S.C. Kernthaler, R. Toumi and J.D. Haigh, Some doubts concerning a link between cosmic ray fluxes and global cloudiness, Geophys. Res. Lett. 26 (1999) 863–865. [13] T.S. Jorgensen and A.W. Hansen, Journal of Atmospheric and Solar-Terrestrial Physics 62 (1997) 73–77. A reply is in the adjacent paper: H. Svensmark and E. Friis-Christensen, Journal of Atmo- spheric and Solar-Terrestrial Physics 62 (1997) 79–80. [14] J.E. Kristjansson and J. Kristiansen, Is there a cosmic ray signal in recent variations in global cloudiness and cloud radiative forcing?, J. Geophys. Res. 105 (2000) 11851–11863. [15] W.B. Rossow, A.W. Walker, D.E. Beuschel, and M.D. Roiter, International Satellite Cloud Clima- tology Project (ISCCP): documentation of new cloud datasets, WMO/TD 737,World Meteorolog- ical Organization, Geneva (1996). [16] M. Lockwood, Changes in galactic cosmic ray flux, solar irradiance and climate over the past 150 years, in Proc. of the Workshop on Ion-Aerosol-Cloud Interactions, ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [17] D.V. Hoyt and K.H. Schatten, The Role of the Sun in Climate Change, Oxford University Press, Oxford (1997). [18] C.E.P. Brooks, Variations in the levels of the central African Lakes Victoria and Albert, Geophysical Memoirs, London, 2 (1923) 327–334. [19] G.W. Lockwood and D.T. Thompson, Photometric variability of Neptune, 1972–2000, in press (2001). [20] N. Marsh and H. Svensmark, GCR and ENSO trends in ISCCP-D2 low cloud properties, submitted to J. Geophys. Res. (2001); also http://www.dsri.dk/∼ndm/CLOUD UPDATE/UPDATE.html. [21] D.L. Hartmann, Radiative effects of clouds on Earth’s climate, in Aerosol-Cloud-Climate Interac- tions,International Geophysics Series, Vol. 54, ed. P.V. Hobbs, Academic Press Inc., San Diego (1993) 151. [22] J. Beer, Ice core data on climate and cosmic ray changes, in Proc. of the Workshop on Ion-Aerosol- Cloud Interactions, ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [23] J.R. Herman and R.A. Goldberg, Sun, Weather, and Climate, NASA SP 426 (1978). [24] K.C. Taylor et al., The Holocene-Younger Dryas transition recorded at Summit, Greenland, Science 278 (1997) 825. [25] W.S. Broecker and G.H. Denton, What drives glacial cycles?, Scientific American 262 (1990) 43– 50. [26] B. Stenni et al.,Anocean reversal during the last deglaciation, Science 293 (2001) 2074–2077. [27] J.P. Sachs et al., Glacial surface temperatures of the Southeast Atlantic Ocean, Science 293 (2001) 2077–2079. [28] R. Muscheler, J. Beer, G. Wagner and R.C. Finkel, Changes in deep-water formation during the Younger Dryas event inferred from 10Be and 14Crecords, Nature 408 (2000) 567–570. [29] H. Renssen, B. van geel, J. van der Plicht and M. Magny, Reduced solar activity as a trigger for the start of the Younger Dryas?, Quaternary International 68Ð71 (2000) 373–383. [30] G.C. Bond and R. Lotti, Iceberg discharges into the North Atlantic on millennial time scales during the last glaciation, Science 267 (1997) 1005–1010. [31] G.C. Bond et al.,Apervasive millennial-scale cycle in North Atlantic Holocene and glacial cli- mates, Science 278 (1997) 1257-1266. [32] A. Ganopolski and S. Rahmstorf, Rapid changes of glacial climate simulated in a coupled climate model, Nature 409 (2001) 153–158. [33] G. Wagner, C. Laj, J. Beer et al.,Reconstruction of the paleo accumulation rate of central Greenland during the last 75 kyr using the cosmogenic radionuclides 36Cl and 10Be, and geomagnetic field intensity data, in press (2001). [34] G.C. Bond, B. Kromer, J. Beer, R. Muscheler, M.N. Evans, W. Showers, S. Hoffmann, R. Lotti- Bond, I. Hajdas and G. Bonani, Persistant solar influence on North Atlantic climate during the Holocene, Science 294 (2001) 2130–2136. [35] M. Magny, Lake level fluctuations in the Jura and French subalpine regions associated with ice- rafting in the North Atlantic and variations in the polar atmospheric circulation, Quaternaire 10 (1999) 61–64. [36] U. Neff et al., Strong coincidence between solar variability and the monsoon in Oman between 9 and 6 kyr ago, Nature 411 (2001) 290–293. [37] B. van Geel, J. Buurman and H.T. Waterbolk, Archaeological and palaeoecological indications of an abrupt climate change in The Netherlands, and evidence for climatological teleconnections, around 2650 BP, J. Quaternary Science 11 (1996) 451–460. [38] M. Grosjean et al.,Alate Holocene (<2600 BC) glacial advance in the south-central Andes (29◦S), northern Chile, The Holocene 8 (1998) 473–479. [39] V.J. Polyak and Y. Asmerom, Late Holocene climate and cultural changes in the southwestern United States, Science 294 (2001) 148–151. [40] P.E. Damon and C.P. Sonett, Solar and terrestrial components of the atmospheric 14Cvariation spectrum, in The Sun in Time,eds. C.P. Sonett, M.S. Giampapa and M.S. Matthews, University of Arizona Press, Tucson (1991) 360. [41] G.A. Brook et al., The Holocene 9 (1999) 695. [42] R.F. Denniston et al., Quat. Res. 53 (2000) 196. [43] D. Verschuren, K. Laird and B. Cumming, Rainfall and drought in equatorial East Africa during the past 1100 years, Nature 403 (2000) 410–414. [44] G. Cini Castagnoli et al., The role of cloud cover on the solar illumination signal recorded by δ13C of a shallow water Ionian Sea core, in Proc. of the Workshop on Ion-Aerosol-Cloud Interactions, ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [45] J. Beer, S. Tobias and N. Weiss, An active Sun throughout the Maunder Minimum, Solar Physics 181 (1998) 237–249. [46] J. Beer, A. Blinov, G. Bonani, H.J. Hofmann and R.C. Finkel, Use of 10Be in polar ice to trace the 11-year cycle of solar activity, Nature 347 (1990) 164–166. [47] M. Lockwood, J. Geophys. Res. 106 (2001) 16021–16038. [48] T. Landscheidt, Solar forcing of El Nino˜ and La Nina,˜ in Proc. of the First Solar & Space Weather Euroconference; The Solar Cycle and Terrestrial Climate, Santa Cruz de Tenerife, Spain, ESA SP-463 (2000) 2135–2140. [49] Climate Change 2001: The Scientific Basis,Intergovernmental Panel on Climate Change, eds. J.T. Houghton et al.,Cambridge University Press, UK (2001). [50] G.C. Reid, Influence of solar variability on global sea surface temperatures, Nature 329 (1987) 142–143. [51] D. Ritson, private communication. [52] W.B. White, J. Lean, D.R. Cayan and M.D. Dettinger, Response of global upper ocean temperature to changing solar irradiance, J. Geophys. Res. 102 C2 (1997) 3255-3266. [53] W.B. White, D.R. Cayan and J. Lean, Global upper ocean heat storage response to radiative forcing from changing solar irradiance and increasing greenhouse gas/aerosol concentrations, J. Geophys. Res. 103 C10 (1998) 21,333–21,366. [54] R.J. Charlson, J.H. Seinfeld, A. Nenes, M. Kulmala, A. Laaksonen and M. Christina Facchini, Reshaping the theory of cloud formation, Science 292 (2001) 2025–2026. [55] Y.I. Stozhkov, N.S. Svirzhevsky and V.S. Makhmutov, Cosmic ray measurements in the atmo- sphere, in Proc. of the Workshop on Ion-Aerosol-Cloud Interactions, ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [56] Y.C. Unruh, S.K. Solanki and M. Fligge, Modelling solar irradiance variations: comparison with observations, including line-ratio variations, in Solar Variability and Climate, eds. E. Friis- Christensen et al., Space Science Reviews 94, Nos. 1–2 (2000) 145–152. [57] J.D. Haigh, The impact of solar variability on climate, Science 272 (1996) 981. [58] D. Shindell et al., Solar cycle variability, ozone, and climate, Science 284 (1999) 305–308. [59] K.J.H. Phillips, Guide to the Sun,Cambridge University Press, UK (1992). [60] N.O. Weiss and S.M. Tobias, Physical causes of solar activity, in Solar Variability and Climate, eds. E. Friis-Christensen et al., Space Science Reviews 94, Nos. 1–2 (2000) 99–112. [61] J. Schou et al.,Helioseismic studies of differential rotation in the solar envelope by the Solar Oscil- lation Investigation using the Michelson Doppler Imager, Astrophysical Journal 505 (1998) 390– 417. [62] E.N. Parker, The passage of energetic charged particles through interplanetary space, Planet. Space Sc. 13 (1965) 9–49. [63] L.J. Gleeson and W.I. Axford, Cosmic rays in the interplanetary medium, Astrophys. Journal 149 (1967) 1115–1118. [64] D.E. Groom et al. (Particle Data Group), The Review of Particle Physics, The European Physical Journal C15 (2000) 150–156. [65] V.I. Ermakov, G.A. Bazilevskaya, P.E. Pokrevsky and Y.I. Stozhkov, Ion balance equation in the atmosphere, J. Geophys. Res. 102 (1997) 23413. [66] S. Tonwar (private communication, based on data from the L3+C experiment at CERN). [67] P.J. Crutzen and F. Arnold, Nitric acid formation in the cold antarctic atmosphere: a major cause for the springtime ‘ozone hole’, Nature 324 (1986) 651. [68] M. Nicolet, On the production of nitric oxide by cosmic rays in the mesosphere and stratosphere, Planet. Space Sci. 23 (1975) 637. [69] C.H. Jackman, J.E. Frederick and R.S. Stolarski, Production of odd nitrogen in the stratosphere and mesosphere: an intercomparison of signal strengths, J. Geophys. Res. 85 C12 (1980) 7495. [70] P.J. Crutzen, I.A.S. Isaksen and G.C. Reid, Solar proton events: stratospheric sources of nitric oxide, Science 189 (1975) 457. [71] A.A. Viggiano and F. Arnold, Ion chemistry and composition of the atmosphere, Handbook of Atmospheric Electrodynamics,Vol. 1, ed. H. Volland, CRC Press, Boca Raton (1998). [72] H. Heitmann and F. Arnold, Composition measurements of tropospheric ions, Nature 306 (1993) 747. [73] A. Krieger and F. Arnold, First composition measurements of stratospheric negative ions and in- ferred gaseous sulphuric acid in the winter arctic vortex: Implications for aerosols and hydroxyl radical formation, Geophys. Res. Lett. 21 (1994) 1259. [74] C.T.R. Wilson, Expansion apparatus, Proc. Royal Society of London A87(1912) 277. [75] J.H. Seinfeld and S.N. Pandis, Atmospheric Chemistry and Physics,Wiley, New York (1998). [76] S. Twomey, Aerosols, clouds, and radiation, Atmos. Environ. 25A (1991) 2435–2442. [77] P.V. Hobbs, Aerosol-cloud interactions, in Aerosol-Cloud-Climate Interactions, International Geo- physics Series, Vol. 54, ed. P.V. Hobbs, Academic Press Inc., San Diego (1993) 33–73. [78] S.E. Schwartz and A. Slingo, Enhanced shortwave cloud radiative forcing due to anthropogenic aerosols, in Clouds, Chemistry and Climate, Proceedings of NATO Advanced Study Workshop, ed. P. Crutzen and V. Ramanathan, Springer, Heidelberg (1996) 191–236. [79] L.F. Radke, J.A. Coakley and M.D. King, Science 246 (1989) 1146–1149. [80] D. Rosenfeld, Suppression of rain and snow by urban and industrial air pollution, Science 287 (2000) 1793. [81] F. Arnold, Multi-ion complexes in the stratosphere: implications for trace gases and aerosols, Na- ture 284 (1980) 610. [82] F. Raes and A. Janssens, Ion-induced aerosol formation in a H2O-H2SO4 system - I. Extension of the classical theory and search for experimental evidence, J. Aerosol Sci. 16 (1985) 217. [83] F. Raes, A. Janssens and R.V. Dingenen, The role of ion-induced aerosol formation in the lower atmosphere, J. Aerosol Sc. 17 (1986) 466. [84] I. Kusaka, Z.G. Wang and J.H. Seinfeld, Ion-induced nucleation: a density functional approach, J. Chem. Phys. 102 (1995) 913–924. [85] I. Kusaka, Z.G. Wang and J.H. Seinfeld, Ion-induced nucleation. II. Polarizable multipolar molecules, J. Chem. Phys. 103 (1995) 8993–9009. [86] R.P. Turco, J.-K. Zhao and F. Yu, A new source of tropospheric aerosols: ion-ion recombination, Geophys. Res. Lett. 25 (1998) 635. [87] F. Yu and R.P. Turco, Ultrafine aerosol formation via ion-mediated nucleation, Geophys. Res. Lett. 27 (2000) 883. [88] M. Kulmala, Atmospheric aerosols: formation and growth, in Proc. of the Workshop on Ion- Aerosol-Cloud Interactions,ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [89] A.D. Clarke et al.,Particle nucleation in the tropical boundary layer and its coupling to marine sulfur sources, Science 282 (1998) 89. [90] R.J. Weber et al., Measurement of new particle formation and ultrafine particle growth at a clean continental site, J. Geophys. Res. 102 (1997) 4375–4385. [91] R.J. Weber et al.,New particle formation in the remote troposphere: a comparison of observations at various sites, Geophys. Res. Lett. 26 (1999) 307–310. [92] M.K. Kulmala, L. Pirjola and J.M. Mak¨ ela,¨ Stable sulphate clusters as a source of new atmospheric particles, Nature 404 (2000) 66. [93] F. Yu and R.P. Turco, From molecular clusters to nanoparticles: The role of ambient ionisation in tropospheric aerosol formation, J. Geophys. Res. 106 (2001) 4797–4814. [94] R. D’Auria and R.P. Turco, A thermodynamic-kinetic model for ionic cluster formation, growth and nucleation, Proc. of the Workshop on Ion-Aerosol-Cloud Interactions, ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [95] F. Yu, Cosmic rays, particle formation, natural variability of global cloudiness, and climate impli- cations, in Proc. of the Workshop on Ion-Aerosol-Cloud Interactions,ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [96] U. Horrak,˜ J. Salm and H. Tammet, Bursts of intermediate ions in atmospheric air, J. Geophysical Research 103 D12 (1998) 13909. [97] J. Bricard et al.,Formation and evolution of nuclei of condensation that appear in air initially free of aerosols, J. Geophysical Research 73 (1968) 4487. [98] K.G. Vohra, M.C. Subba Ramu and T.S. Muraleedharan, An experimental study of the role of radon and its daughter products in the conversion of sulphur dioxide into aerosol particles in the atmosphere, Atmospheric Environment 18 (1984) 1653. [99] B. Zuberi et al., Heterogeneous freezing of aqueous particles induced by crystallized (NH4)2SO4, ice, and letovicite, J. Physical Chemistry A 105 (2001) 6458–6464. [100] B.A. Tinsley and G.W. Dean, Apparent tropospheric response to MeV–GeV particle flux varia- tions: a connection via the solar wind, atmospheric electricity and cloud microphysics, J. Geophys- ical Research 96 (1991) 22283. [101] B.A. Tinsley, Influence of solar wind on the global electric circuit, and inferred effects on cloud microphysics, temperature, and dynamics in the troposphere, in Solar Variability and Climate,eds. E. Friis-Christensen et al., Space Science Reviews 94, Nos. 1–2 (2000) 231–258. [102] H.R. Pruppacher and J.D. Klett, Microphysics of Clouds and Precipitation, 2nd ed., Kluwer Aca- demic Publishers, The Netherlands (1997). [103] R.J. Anderson, R.C. Miller, J.L. Kassner and D.E. Hagen, A study of homogeneous condensation- freezing nucleation of small water droplets in an expansion cloud chamber, J. Atmospheric Sciences 37,No. 11 (1980) 2508–2520. [104] A.E. Waibel, T. Peter, K.S. Carslaw, H. Oelhaf, G. Wetzel, P.J. Crutzen, U. Poschl,¨ A. Tsias, E. Reimer and H. Fischer, Arctic ozone loss due to denitrification, Science 283 (1999) 2064–2069. [105] D.R. MacGorman and W.D. Rust, The Electrical Nature of Storms, Oxford University Press, UK (1998). [106] K. Schlegel et al., Thunderstorms, lightning and solar activity—Middle Europe, J. Atmospheric and Solar-Terrestrial Physics, 63 (2001) 1705–1713. [107] R.G. Roble, On solar-terrestrial relationships in the atmospheric electricity, J. Geophys. Res. 90 (1985) 6000. [108] R.E. Orville and G.R. Huffines, Lightning ground flash measurements over the contiguous United States: a ten-year summary 1989–1998, in Proc. 11th Int. Conf. on Atmospheric Electricity, Al- abama, USA (1999) 412. [109] B.J. Kellett, The production of atmospheric nitric oxide by cosmic rays and solar energetic par- ticles, in Proc. of the Workshop on Ion-Aerosol-Cloud Interactions, ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [110] Yu.I.Stozhkov et al.,Rainfalls during great Forbush decreases, Il Nuovo Cimento 18C (1995) 335. [111] M.I. Pudovkin and S.V. Veretenenko, Cloudiness decreases associated with Forbush decreases of galactic cosmic rays, J. Atmos. Solar-Terrestrial Phys. 75 (1995) 1349. [112] A.I. Rusanov and F.M. Kuni, Reformulation of the thermodynamic theory of nucleation on charged particles, J. Colloid and Interface Science 100 (1984) 264. [113] A.V. Gurevich, K.P. Zybin and R. Roussel-Dupre, Lightning initiation by simultaneous effect of runaway breakdown and cosmic ray showers, Phys. Lett. A 254 (1999) 79–87. [114] A.P. Chubenko et al., Intensive X-ray emission bursts during thunderstorms, Phys. Lett. A 275 (2000) 90–100. [115] C.B. Moore, K.B. Eack, G.D. Aulich and W. Rison, Energetic radiation associated with lightning stepped-leaders, Geophys. Res. Lett. 28 No. 11 (2001) 2141–2144. [116] P. Galison, Image and Logic; a Material Culture of Microphysics, University of Chicago Press, Chicago (1997). [117] A. Nenes, P.Y. Chuang, R.C. Flagan and J.H. Seinfeld, A theoretical analysis of cloud condensa- tion nucleus (CCN) instruments, J. Geophys. Res. 106 D4 (2001) 3449–3474. [118] P.E. Wagner, Experiments on nucleation processes in aerosols, in Proc. of the Workshop on Ion- Aerosol-Cloud Interactions,ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [119] J.L. Schmitt, Comments on the operation of a Wilson expansion cloud chamber, in Proc. of the Workshop on Ion-Aerosol-Cloud Interactions,ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [120] N.N. Das Gupta and S.K. Ghosh, A report on the Wilson cloud chamber and its application in physics, Rev. Modern Physics 18 (1946) 225–290. [121] H. Bunz et al., The novel aerosol chamber facility AIDA: status and first results, in Proc. 7th Eu- ropean Symposium on the Physico-Chemical Behaviour of Atmospheric Pollutants, eds. B. Larson, B. Versino and G. Angeletti, Venice (1996) 673. [122] O. Mohler¨ et al.,The Karlsruhe aerosol chamber facility AIDA: Technical description and first results of homogeneous and heterogeneous ice nucleation experiments, in Proc. of the Workshop on Ion-Aerosol-Cloud Interactions,ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [123] G.B. Bowden, private communication. [124] A. Herve´ and G. Bachy, Huge electrohydraulic servovalves control 2-ton cryogenic piston, Hy- draulics & Pneumatics, Pention/IPC, Cleveland, Ohio (1977); and, The hydraulic expansion servo- system of the Big European Bubble Chamber, CERN preprint CERN/EF 77-6 (1977). [125] Bellofram Corporation, Burlington, MA 01803, USA, http://www.marshbellowfram.com. [126] W.C. Hinds, Aerosol Technology: Properties, Behavior, and Measurement of Airborne Particles, John Wiley & Sons, Inc., New York (1998). [127] P.E. Wagner, A constant-angle Mie scattering method (CAMS) for investigation of particle forma- tion processes, J. Colloid Interface Sci. 105 (1985) 456. [128] F. Pavese and G. Molinar, Modern Gas-Based Temperature and Pressure Measurements, The In- ternational Cryogenics Monograph Series, Plenum Press, New York (1992). [129] T.P. Burke and J.A. Scott, The production of condensation nuclei by alpha radiation, Proc. Royal Irish Academy 73 A (1973) 151–158. [130] J.M. Makel¨ a,¨ Studies on irradiation induced aerosol particle formation processes in air containing sulphur dioxide, Ph.D. thesis, Univ. Helsinki, Acta Polytechnica Scandinavica, Applied Physics Series 82 (1992). [131] C.F. Clement and R.G. Harrison, Enhanced localised charging of radioactive aerosols, J. Aerosol. Sci. 31,3(2000) 363–378. [132] A.S. Wexler and Z. Ge, Hydrophobic particles can activate at lower relative humidity than slightly hygroscopic ones: a Kohler¨ theory incorporating surface fixed charge, J. Geophys. Res. 103 D6 (1998) 6083–6088. [133] M. Kulmala, A. Laaksonen, R.J. Charlson and P. Korhonen, Clouds without supersaturation, Na- ture 388 (1997) 336. [134] P.Y. Chuang, R.J. Charlson and J.H. Seinfeld, Kinetic limitations on droplet formation in clouds, Nature 390 (1997) 594–596. [135] A.D. Erlykin and A.W. Wolfendale, Cosmic rays and the Earth’s atmosphere, in Proc. of the Work- shop on Ion-Aerosol-Cloud Interactions, ed. J. Kirkby, CERN, Geneva, CERN 2001-007 (2001). [136] A.K. Barlow and J. Latham, A laboratory study of the scavenging of sub-micron aerosol by charged raindrops, Quart. J. R. Met. Soc. 109 (1983) 763–770. [137] G.R. Evans and A.A. Watson, Production of hypersensitive condensation nuclei, Nature 196 (1962) 729; G.R. Evans et al., Some remarks on the origin and elimination of background cloud in expansion cloud chambers, J. Sci. Inst. 41 (1964) 770; G.R. Evans and A.A. Watson, Nitrogen dioxide vapour and the creation of condensation nuclei, Proc. Royal Society of Edinburgh LXVII part III-10 (1966) 136. [138] E.N. Parker, Summary and perspectives, in Proc. of the First Solar & Space Weather Euroconfer- ence; The Solar Cycle and Terrestrial Climate, Santa Cruz de Tenerife, Spain, ESA SP-463 (2000) 263–269. CONCLUSIONS OF THE WORKSHOP ON ION-AEROSOL-CLOUD INTERACTIONS, CERN, 18-20 APRIL 2001 A.W. Wolfendale Department of Physics, University of Durham, South Road, Durham, DH1 3LE, U.K. Abstract Recent observations suggest that cosmic rays may play a significant role in the climate. In particular, satellite data have revealed a surprising correlation between cosmic ray intensity and the fraction of the Earth covered by low clouds. Since the cosmic ray intensity is modulated by the solar wind, this could provide an important clue to the long-sought mechanism connecting solar and climate variability. Moreover, if this connection were to be established, it could have significant consequences for our understanding of the solar contributions to the present global warming, since the cosmic ray intensity has fallen during the 20th century due to a more-than-doubling of the strength of the solar wind. In order to the test whether cosmic rays and clouds are causally linked and, if so, to understand the microphysical mechanisms, a novel experiment known as CLOUD has been proposed at CERN by an interdisciplinary collaboration of atmospheric, solar-terrestrial, cosmic ray and particle physicists. CLOUD proposes to use a CERN pion beam as an artificial source of cosmic rays. The beam would pass through an expansion cloud chamber in which the atmospheric conditions within clouds throughout the atmosphere could be reproduced. All parameters of the experiment would be precisely controlled and measured. A workshop was recently held at CERN to discuss the scientific case for a connection between cosmic rays and clouds, and to review the proposed CLOUD facility. The outcome was a clear consensus that the scientific indications of a cosmic ray-cloud link are both interesting and important, and that plausible microphysical mechanisms exist but their significance is not yet known. There was unanimous agreement on the urgent need to perform controlled laboratory measurements to test the cosmic ray-cloud link in a particle beam at CERN, as proposed by the CLOUD experiment. Further details on the outcome of the workshop are provided below. 1. INTRODUCTION The European Geophysical Society, the European Physical Society and the European Science Foundation co-sponsored an inter-disciplinary "Workshop on Ion-Aerosol-Cloud Interactions" at CERN, 18-20 April 2001. The workshop was attended by about 50 physicists representing 14 countries and drawn from the atmospheric, aerosol, palaeoclimatological, solar-terrestrial, cosmic ray and particle physics communities. This aim of the meeting was twofold: 1) To review the present knowledge of ion-aerosol-cloud interactions and their possible role in solar- climate variability. 2) To review the proposed CLOUD Atmospheric Research Facility using a particle beam at CERN. 2. PROCEDURE In order to arrive at a consensus on the workshop conclusions, the meeting closed with a "Workshop Summary Panel" discussion. The panel members comprised: Sir Arnold Wolfendale /University of Durham (chairman) Maurice Jacob /Chairman of the Joint Astrophysics Division of EAS and EPS Mike Lockwood /RAL, President of Solar-Terrestrial Sciences, EGS Richard Turco /University of California, Los Angeles Paul Wagner /University of Vienna To focus the discussion, the chairman presented four basic (but inevitably over-simplified) questions to the panel and then to the floor: 1) "Does cosmic ray ionization play a role in the climate?" 2) "Is the mechanism: ionization -> aerosol -> cloud understood?" 3) "Is the case (scientific motivation) for a cosmic ray influence on cloud cover agreed?" 4) "Would the CERN 'CLOUD' facility satisfy a need?" As well as a general discussion, the chairman invited all those present (numbering about 30 in the final session) to vote either "No", "?" or "Yes" to each question. Finally, after each question had been discussed and voted in turn, the chairman invited the panel and floor to express their opinions on a fifth question: 5) "Why at CERN?" The points raised during the discussion on each question are summarised below, together with the results of the voting. 3. RESPONSES TO THE QUESTIONS 3.1 "Does cosmic ray ionization play a role in the climate?" Elaboration: This question asked whether cosmic rays have the potential to affect the climate and whether there is evidence that it may be happening. The question did not ask whether cosmic rays do indeed significantly affect the climate - which is clearly unanswerable at present. Discussion: Clear historical correlations of sunspot/solar variability and changes in the Earth's climate were presented at the workshop. For example, Beer, Lockwood and van Geel showed examples such as the Maunder Minimum (Little Ice Age), the circa-850BC climate anomaly and the Younger Dryas cold event at the end of the last Ice Age (12.9-11.6 kyear before present) which are associated with solar variability as revealed in the sunspot record and in the cosmogenic isotope record in ice-cores (10Be) and tree-rings (14C). These data directly indicate the prevailing galactic cosmic ray (GCR) intensities, which are modulated by the solar wind (and by slower changes in the Earth's magnetic dipole). However the solar/GCR-climate correlations are sometimes present and sometimes not. This may reflect the complexity of the Earth's climate - that many factors are important and they interact in a complex way. The climate may have stable states such that a correlation may persist for some decades and then disappear for a while. In addition, whatever caused those earlier natural climate shifts may also be interacting with today's anthropogenic contributions in the atmosphere to produce a yet more complex response, for example anthropogenic sulphur dioxide and its effect on cloud cover. However, correlations do not demonstrate cause and effect, so the present data are unable to separate whether the Sun-Earth coupling is via electromagnetic radiation (total irradiance/UV/...) and/or via energetic cosmic rays (galactic/solar). But it is important to note that these are the only possible vectors (it is unlikely that the solar wind itself could be directly responsible) - and so at least one of them must be implicated. In the case of cosmic rays one should in particular study and understand the amplification factors that would be necessary to enhance their role despite their very small energy input (roughly equivalent to that of starlight) in comparison with total solar irradiance. (The vast disparity of energies, by itself, does not exclude the possibility of an effect; there are numerous examples in physics of large energy amplification factors, such as a nuclear chain reaction released by a few initial neutrons.) In the case of the current global warming, there is increasing agreement that the climate model fits to the temperature record need to amplify the solar contribution by about a factor 3. The presently-assumed solar contribution is only from the (Lean et al., 1995) direct irradiance changes. An additional, indirect, solar contribution could either decrease or increase the projections of the anthropogenic effects. (The latter possibility arises since an increased solar attribution during the last century could indicate a steeper anthropogenic rise in recent decades.) The satellite data analysis presented at the workshop by Svensmark indicates a solar cycle correlation with low cloud cover, suggesting that the solar-climate mechanism may involve clouds. Again, at this stage both electromagnetic radiation and GCRs remain as candidates. This may provide the first clue to the long-sought amplification mechanism linking solar and climate variability. However the underlying processes may involve subtleties since the observed solar correlation is confined to low clouds, and the global correlation map of low cloud cover shows no preference for high geomagnetic latitudes - both of which appear to be counter-intuitive at first sight. Vote: The distribution of votes on the question "Does cosmic ray ionization play a role in the climate?" was equally divided between "?" and "Yes", with zero votes for "No". This implies that there are reasonable indications that cosmic rays have the potential to affect the climate but that the question of whether they are significant is far from settled. 3.2 "Is the mechanism: ionization -> aerosol -> cloud understood?" Elaboration: This question asked whether there is any microphysical understanding of the mechanism(s) by which cosmic ray ionization could affect 1) the nucleation of new aerosol and 2) the lifetime, albedo or other properties of clouds. Discussion: There is now strong evidence to support the existence of the first step. Yu and Turco presented the results of their theoretical studies of ion-induced nucleation and conclude that ions play an important role in the creation and early growth of ultrafine condensation nuclei (UCN) from trace vapours such as sulphuric acid. These frequently occur in clean environments (such as over oceans) at very low concentrations where classical nucleation theory predicts no nucleation should occur - but nevertheless nucleation is observed. Yu and Turco find that that the presence of charge serves to stabilise the embryonic clusters, and their ion-mediated model agrees with the experimental observations. In addition, Yu reported on the effect of variations of GCR ionization at different altitudes and concludes that it can be the limiting factor to aerosol nucleation at low altitudes, whereas at high altitudes, where the ionization rate is up to a factor 20 larger, other parameters such the trace gas concentrations become the limiting factor. This would provide a possible explanation why the solar modulation is observed only in low clouds. As well as theoretical developments, F.Arnold presented at the workshop the first direct observation of ion-induced nucleation in the laboratory, and also aircraft measurements of the ion mass spectrum in the atmosphere which extend to large ions and indicate the presence of ion-induced nucleation. These theoretical and experimental developments represent significant progress and lay to rest a common criticism raised against the cosmic ray-cloud hypothesis - namely that no microphysical mechanism exists to connect cosmic rays to clouds. At least one mechanism exists but whether or not it is significant is not yet known. Whereas there now seems little doubt that cosmic rays can influence the nucleation of trace condensable vapours under certain conditions, the effect of these extra UCN on the cloud condensation nuclei (CCN) that seed cloud droplets is an open question. Equally, the influence of GCRs on the growth process of other aerosols or on the activation of CCN into droplets is not known. However, if cosmic rays could indeed modify the CCN number concentration in certain regions of the atmosphere then this may affect both cloud lifetimes and albedo. Furthermore, GCRs may have other effects on clouds such as the electrofreezing of supercooled liquid droplets, influences on the global electrical circuit and electric field strength, and the production of trace reactive chemicals (NO, OH) which could affect atmospheric chemistry at certain altitudes. In summary, there are now actually several mechanisms that have been identified by which GCRs may potentially affect clouds, but they are yet to be investigated experimentally and quantified. Vote: The distribution of votes on the question "Is the mechanism: ionization -> aerosol -> cloud understood?" was bimodal. There was a 100% "Yes" vote for the first step, indicating that at least one mechanism is explicable theoretically, if not proven experimentally (although the first direct observations of ion-induced aerosol formation were presented at the workshop). However whether or not these UCN have a significant effect on CCN is essentially unknown. This was reflected in the vote for the second step which was equally divided between "No" and "?", with zero votes for "Yes". This latter vote indicates also the poor experimental and theoretical understanding of the effects of ionization on the aerosol growth and activation processes, and on other areas where it may play a role. 3.3 "Is the case (scientific motivation) for a cosmic ray influence on cloud cover agreed?" Elaboration: This question asked whether the scientific indications are sufficiently interesting and important to merit a controlled laboratory experiment on the influence of cosmic rays on clouds. Discussion: In view of the preceding discussion on the first two questions, there was little extra discussion before a vote was taken on this third question. However it was pointed out that the GCR- cloud hypothesis may be the very first hard clue we have as to what is behind the solar-climate correlations that have been observed over the last two centuries. If our only tool is correlations, we may continue another two centuries and still not be able to understand the underlying mechanism. However at last we have a definite hypothesis that can be tested experimentally: "Are cosmic rays affecting cloud formation?". The question is so important that we should pursue it. Vote: On the question "Is the case (scientific motivation) for a cosmic ray influence on cloud cover agreed?": 100% "Yes". 3.4 "Would the CERN 'CLOUD' facility satisfy a need?" Elaboration: This question asked whether the CERN 'CLOUD' facility would provide important new experimental data on the subject of ion-aerosol-cloud interactions and whether the facility would be complementary to other experiments in this field. Discussion: It was agreed that the CLOUD facility is timely for several reasons. First of course are the recent satellite observations of a solar modulation of cloud cover, and its possible effect on the climate and global warming. Recent theoretical progress has been made on the understanding of ion- mediated effects on aerosols by Yu, Turco, Okuyama and others, and there is now an urgent need for experimental data to test these models. Finally, it is only in the last few years that the necessary precision experimental tools have been developed which will allow the proposed CLOUD experiments to be carried out. At the workshop Möhler presented the experience with the Karlsruhe aerosol chamber facility, AIDA, which has successfully demonstrated the feasibility of aerosol growth experiments in the laboratory under atmospheric conditions. Furthermore Wagner described recent measurements which show expansion cloud chambers to be versatile and high-precision experimental tools that are ideally suited for the proposed studies. With expansion cloud chambers, well-defined thermodynamic conditions can be produced over large volumes and, with the use of a CERN particle beam, the cosmic ray conditions throughout the atmosphere can be recreated. The proposed CLOUD facility would be the world's first to precisely simulate the conditions inside clouds at all altitudes and latitudes, and to investigate the effects of ionizing particle radiation on aerosol and cloud processes. In addition to aerosol nucleation, growth and activation experiments, CLOUD will be able to measure the effect of cosmic ray ionization on a wide range of atmospheric processes. For example, Carslaw described at the workshop how ionization has been proposed as the possible mechanism by which polar stratospheric clouds freeze. Discovery of the freezing mechanism in these clouds is crucial to our understanding of de-nitrification and subsequent ozone loss over the poles. Kellett showed evidence for production of nitric oxide in the atmosphere by energetic solar cosmic ray events and suggested that GCRs may affect the rate of NO production in the lower atmosphere by affecting lightning production. Stozhkov in fact presented ground-based data collected in some regions of the United States that shows a correlation between the GCR intensity and the frequency of lightning. Stozhkov also suggested that a preferential activation of water droplets on negative ions may be responsible for charge separation in clouds, and therefore lightning. He also presented data indicating a decreased rainfall during Forbush decreases, and increased rainfall during energetic solar cosmic ray events. GCRs are responsible for the fair weather ionization throughout most of the lower atmosphere and are therefore a key element in the global electrical circuit. Harrison summarised several atmospheric electricity processes, such as electrofreezing, aerosol charging, and the scavenging of charged aerosols by cloud droplets, that may play important roles in cloud microphysics. The CLOUD facility can investigate the fundamental physics that underlies each of these processes. It would provide important microphysical measurements to help the interpretation of atmospheric observations by programmes such as the ESF SPECIAL network to study Sun-Earth links, which Rycroft described at the meeting. In short, there is not a single "need" for CLOUD but, rather, a wide range of "needs", making the concept of a facility appropriate for the project. Furthermore, CLOUD should be seen as providing an essential and complementary contribution in support of an extensive on-going solar-terrestrial experimental programme involving satellites and ground-based stations. Vote: On the question "Would the CERN 'CLOUD' facility satisfy a need?": 100% "Yes". 3.5 "Why at CERN?" Elaboration: This question asked why the CLOUD facility should be located at CERN. Discussion: There are two basic reasons why CERN is uniquely suitable for the CLOUD facility: a) the particle beam and b) technological expertise and excellence in the equipment needed for the experiment, together with rapidly-increasing knowledge by talented staff of the detailed research problems to be addressed. The theoretical studies of Yu, Turco and others have shown that ionization effects are highly non-linear and so experiments must reproduce ionization rates and ionization densities (dE/dx) close to natural GCRs. Such measurements have so far never been achieved with radioactive sources despite experiments over the last 40 years. However, a CERN pion beam closely duplicates natural GCRs and provides a precisely controlled and delivered particle ionization inside the active volumes of the experiment. To answer the scientific questions addressed by CLOUD requires a sophisticated and technically-challenging experimental apparatus - one that is beyond the capabilities of individual institutes but well within the scope of the experiments for which CERN is well known. In particular CERN has key expertise in the expansion cloud chamber, from its experience with BEBC and other bubble chambers. In this sense the CLOUD facility could be considered as a "technology transfer" from CERN, but to another research community rather than to industry, and on a subject of great interest to society. The project represents a unique interface that brings together cosmic ray/particle physics - which is within the mandate of CERN - and atmospheric physics. Such a facility may attract EU funding support. The facility coincides with a research hiatus at CERN over the next 5 years while the LHC is being constructed. As well as issues related to the beam and technological expertise, the CLOUD facility has attracted an enthusiastic interdisciplinary collaboration with an unprecedented range of experience and skills. However, the prime reasons "Why at CERN?" are the importance of the CLOUD facility to science and to society; and CERN is the unique European host. Summary of the Conclusions of the Workshop on Ion-Aerosol-Cloud Interactions, CERN, 18-20 April 2001 Voting result 1) "Does cosmic ray ionization play a role in the climate?" No ? Ye s 2) "Is the mechanism understood for: a) ionization -> aerosol? No ? Ye s and b) aerosol -> cloud?" No ? Ye s 3) "Is the scientific motivation for a cosmic ray influence on cloud cover agreed?" No ? Ye s 4) "Would the CERN 'CLOUD' facility satisfy a need?" No ? Ye s Fig.1 Summary of the voting results at the workshop. = 11.1 ± 2.1 fSm-1, compared to day 170 when -1 = 14.8 ± 8.6 fSm . There was also a weak positive correlation (significant to > 90%) with Sg, suggesting that there was a different régime to the solar-induced turbulent mixing discussed above. Mechanical turbulence may have dominated, particularly in the morning when the wind was approximately constant at 3.5 ms-1, due to the remnants of a cold front. On day 170 the wind arises primarily from local convective mixing and shows the usual diurnal cycle with a maximum in the afternoon. During the afternoon, which was sunnier on day 165, both the s and wind traces for days 165 and 170 are more closely related, with the conductivity minimum occurring at similar times, associated with the maximum temperature. 2.4 Factors influencing conductivity for two cloud-free periods Here the predominantly clear period during day 169 is compared to the same time for day 170. Direct comparisons of two days are not normally possible because of the great variation in meteorological conditions. Days 169 and 170 were therefore rare because the synoptic conditions and incoming solar radiation were very similar, allowing detailed analysis of specific effects on s. Day 169 was almost cloud-free from 0530 until sunset. Some cirrus cloud was recorded manually at 1000, radar data† however showed no cloud over the whole day. There appears to have been – st. dev. (fSm-1) (ms ) RH(%) Dirn (deg)