Met Office College - Course Notes Cloud physics Contents 1. The development of cloud physics 2. Aerosol spectra 3. Droplet formation 4. Rate of growth by condensation 5. Growth by collision-coalescence 6. Ice processes 7. Drop size distributions 8. Cloud seeding 9. Lightning – cloud charging mechanisms 10. References and further reading Crown Copyright. Permission to quote from this document must be obtained from The Principal, Met Office College, FitzRoy Road, Exeter, Devon. EX1 3PB. UK. Page 1 of 27 Last saved date: 9 March 2016 FILE: MS-TRAIN-COLLEGE-WORK-D:\533580417.DOC Met Office College “So the cloudlets grow by mutual fusion and scud before the winds, until the time comes when a raging storm arises”. Lucretius 1st century BC 1. The development of cloud physics Cloud-seeding studies in the USA and Germany in the late 1930’s and 1940’s highlighted the dearth of understanding of cloud precipitation mechanisms. By the 1960’s cloud physics laboratories were investigating basic ‘cloud physics’ The developing requirements of Numerical Weather Prediction intensified the need to understand, and parametrise, cloud physics processes. Studies have been increasingly complemented by fieldwork with instrumented aircraft, radar and aided by satellite imagery. 1.1 The cloud-modelling problem The atmosphere produces precipitation, both a sink of moisture in the atmosphere and a source at the surface. An NWP model must remove moisture from its grid in order to simulate this process. The scheme ‘has knowledge’ of temperature, water vapour, liquid and ice content centred in each grid box. The model must estimate changes in these quantities due to precipitation processes. There are many physical processes important in clouds (see selection below); the rôle of the ice-phase is being increasingly appreciated. Table 1 Some important cloud physics processes/properties Droplets Ice phase Hail Raindrops Activation Nucleation Dry hail growth Melting Condensation Deposition – growth, sublimation Wet hail growth Collision with ice or droplets Fall speed Collision & coalescence Spectrum evolution Droplet shedding Shape Fall speed and mode Ventilation Collision, aggregation Collection of droplets (riming) Splintering, Page 2 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Nucleation of supercooled drops Evaporation of raindrops Fall speed, ventilation Drop break-up For all processes: The influence of Cloud physics multiplication electric fields Parametrisation means processes must be simplified – the scheme must capture the necessary physics from a knowledge of only a few variables. A suitable ‘scheme’ is shown in Fig. 1. Thus a cloud droplet is assumed to respond instantaneously to changes in the model background humidity – the fine details of the Kelvin, Solute and Köhler curves are irrelevant! (Wilson and Ballard, 1999). Processes leading to precipitation formation Fig. 1 Parametrisation needs an in-depth appreciation of the processes listed in Table 1 and Fig.1 to establish which ones are less/more important for the modeller. There are still many important and exciting discoveries to be made in cloud physics, from the evolution of the droplet spectrum to the nature of cloud charging mechanisms. Some of these problems have been occupying cloud physicists for over half a century! 2. Aerosol spectra Aitken in Scoltand and Coulier in France, working in the 1880’s, established the fundamental rôle of condensation nuclei for cloud droplet development; Aitken worked on the nature of aerosols in Page 3 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College different air masses for over 30 years. Fig.2 contrasts the cloud droplet spectra in continental and maritime air masses. The German investigator, Junge, working in the 1950s, found that particles of r > 0.1m obeyed a size distribution law of the form: n(r ) dN A d (log r ) r where A is a constant and dN = n(r) d (log r) is the number of particles per cm-3 in the radius interval (log r). Furthermore Junge found to be typically about 3. This implies that the sum of the cubes of radii for particles in each logarithmic interval of radius is equal. In other words the contribution to the total aerosol mass from particles between 0.1 and 1m radius is similar to that from particles of between 1m and 10m Junge: n( r ) dN A d log r r Junge : n ( r ) A dN d log r r dN n ( r ) d (log r ) : particles cm 3 in radius interval :d (log r ) • ~3 so r3 in each log r interval is equal • Mass contribution 0.1 to 1 similar to that for1 to 10 } • 0.1 to 1: accumulation mode; 10 to 20 coarse particle mode • encouraged by Brownian aggregation; and by gravitational settling radius (Fig 3). Fig.2 Fig.3 Fig. 3 shows size distributions of aerosols collected over Miles City, Montana, the ‘envelopes’ containing many different samples. As well as particle number, there are distributions based on surface area: dS/d(log d) and volume: dV/d(logd). The Junge distribution, which would appear as a straight line in Fig. 3, illustration (a), only provides a rough fit to the data which show distinct inflexion points. Furthermore, if = 3 provided a good fit to these data, Page 4 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics the trace in Fig. 3(c) would be nearly horizontal. The maximum between 0.1 and 1m, especially noticeable in Figs.3(b) and 3(c), is known as the accumulation mode, which is explained by the tendency of particles smaller than 0.lm to collide with one another due to Brownian motion and coagulate. Between 10 and 20m exists another maximum known as the coarse particle mode, which is dominated by dust, combustion products and sea salt. This has been shown to depend on wind speed and distance from the sources of the aerosols. The reason for its occurrence is the tendency for gravitational settling to remove larger particles. It is interesting to note that it is the large aerosols (0.1 – 1.0m) which contribute most to the total aerosol surface area. 3. Droplet formation 3.1 The adiabatic cloud liquid water content. A first approximation to the liquid water content of a sample of cloudy air can be made by assuming adiabatic ascent of moist air. The mass of liquid water is given by the initial humidity mixing ratio of the air parcel, (rso), minus the current saturation humidity mixing ratio, (rs), taking account of cooling at the Dry Adiabatic Lapse Rate while the air is unsaturated and Saturated ALR while the air is saturated: q l = rso - rs This calculation can be made on a thermodynamic diagram, such as a tephigram, and converted into liquid water per unit volume rather than unit mass by multiplying by air density, or p/ RT. In practice, however, actual values of cloud liquid water content are lower than the adiabatic value, suggesting that mixing of drier air from outside the cloud occurs to a significant degree. 3.2 The Kelvin (curvature) effect The equilibrium vapour pressure over a curved surface of water is greater than that over a plane surface. This can be explained in terms of surface tension or in terms of molecular attraction. At a curved surface there is a weaker net attraction holding water molecules in the liquid mass, since each molecule at the surface is more exposed. Consequently more molecules escape into the vapour phase than over a plane surface and so the vapour pressure exerted is greater. This pressure is given by: 2 e s (r ) e s () exp rR v L T …(1) Page 5 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College where e s() is the equilibrium vapour pressure over a plane surface of water, is the surface tension, r the radius of curvature, Rv the gas constant for water vapour and L the density of water. As a droplet's size increases, the vapour pressure exerted decreases, and liquids with higher values of surface tension (and therefore greater molecular attraction) show a proportionally greater increase in vapour pressure over a curved surface. While e > e s(r), a droplet will grow by diffusion of water vapour, i.e. condensation, whereas it will shrink by evaporation if e <e s(r). A critical radius, r c may be defined for a given vapour pressure by writing e = e s(r) in equation (1).This defines the size of a droplet that will remain in equilibrium for a given vapour pressure. Setting e/e s() = S, a measure of relative humidity: rc 2 Rv LT ln S Thus high supersaturations are required for very small droplets to remain stable. For S = 1.01 (i.e. 1% supersaturation), r c = 0.121m at 00C, while for spontaneous (homogeneous) nucleation S must be very large (~ 6 to 8) since chance aggregations of water molecules form only tiny droplets. 3.3 The solute effect Molecules of solute at the surface of a droplet reduce the number of escaping water molecules by occupying sites on the water surface, thus lowering the vapour pressure. For a plane water surface, the reduction in vapour pressure due to solute is given by: n0 e' n 1 e s () n n 0 n0 (2) where e' is the equilibrium vapour pressure over the solution, n 0 is the number of molecules of water and n is the number of molecules of solute (n<<n 0). If the molecules are dissociated into ions, as is the case with a salt, n must be multiplied by a factor, i, which represents the degree of ionic dissociation. i = 2 is considered appropriate for many calculations. Introducing a constant, b, equation (2) simplifies to (Mason, 1971): e s' (r ) b 1 3 e s () r Combining this result with Kelvin's equation (1), Page 6 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics e s' ( r ) b 1 exp e s ( ) r 3 a r where: a = 2 / R vLT Providing r is not too small, a good approximation to this equation for realistic values of a and b is: S e s' (r ) a b 1 3 e s () r r (3) 3.4 The Köhler curve To find the maximum value of S in equation (3), in other words the greatest value of equilibrium relative humidity for a droplet of given solute concentration, dS/dr = 0, i.e. - a/r2 + 3b/r4 = 0. This gives: r* 3b a Substituting into the expression for relative humidity: S* 1 4a 3 27 b So the maximum equilibrium relative humidity is greater than 100%, the margin of supersaturation depending on the temperature, density, • Equilibrium saturation ratio of a solution droplet formed on Ammonium Sulphate nucleus, mass 10-18 kg. surface tension (factor a), and the solution concentration (factor b) Page 7 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College Fig. 4 Fig. 4 shows a Köhler curve for a water droplet growing on a salt nucleus of ammonium sulphate of mass 10-18 kg. When the radius is small the solute effect is important, while there is a fairly rapid transfer in dominance to the curvature effect. In general, at droplet radii above ~5m, neither effect is significant, i.e. the equilibrium relative humidity approaches 100%, (S=1). Droplets of r < r* are termed haze droplets, remaining in stable equilibrium for a given RH, whilst droplets passing r* in radius become unstable and will continue to grow, reducing the environmental supersaturation and causing the haze droplets to shrink. Table 2 shows examples of critical supersaturations for droplets formed on sodium chloride. Table 2: Values of r* and S* as functions of NaCl nucleus concentrations at 273oK Mass of dissolved salt (kg) r* (m) S*-1 (%) 10-18 0.0223 0.42 10-17 0.0479 0.13 10-16 0.103 0.042 10-15 0.223 0.013 10-14 0.479 0.0042 3.5 "Cloud processing" of aerosols droplets; The Met Office’s Met Research Flight results (Osborne, 1996) suggest that various processes, acting during the condensation/evaporation cycle of cloud droplets, can result in residual aerosols which are highly effective as cloudy condensation nuclei , CCN, Figs. 5 and 6. Processes are such that: a) aerosol mass within a droplet can be increased by absorbing SO2. Ionisation and oxidation leads to sulphate ion formation and, ultimately, a larger more effective CCN; b) aerosols too small to be CCN initially can be scavenged to increase the aerosol mass; c) coalescence of cloud drops also results in larger residual potential CCN material; Page 8 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics d) entrainment from above the cloud layer can result in a local, sparser, population of larger drops; e) dimethylsulphide (DMS) from the air-sea interface is thought to form CCN by homogeneous nucleation. Thus increases in DMS are thought to Ship’s plume aerosol size distributions, AVHRR imagery 27/6/87, western seaboard of USA. Ships’ tracks in low level cloud. Fig.5 • High levels of anthropogenic CCN are provided to Sc-topped boundary layer. Aerosol spectra, measured by C-130, within plume near cloud base. • Pronounced bulge after 120km with local max at 0.3m - due to ‘cloud processing’. These aerosols likely to be good CCN and effective at scattering solar radiation (Osborne 1996). yield greater cloud albedoes through increased CCN concentrations. Fig. 5 Fig.6 The effect on solar scattering characteristics of the cloud (susceptibility) have implications for parametrisation, climate change estimates etc. 3.5.1 Cloud susceptibility Cloud susceptibility describes the sensitivity of cloud top albedo, A, over the visible wavelength band, to changes in cloud droplet concentration, Nd (Osborne, 1996): Susceptibility, S dA dN d S can be related to aerosol concentration; studies have concentrated on Sc cloud sheets. The implications of possible anthropogenic aerosol impact on cloud droplet concentration, and hence on the radiation balance of Sc layers (and thus on climate modelling), are clear. 4. Rate of growth by condensation Consider a water drop of mass m and radius r growing by slow diffusion of water vapour. A steady state is achieved with the water vapour flux independent of time and distance, R, from the centre of the droplet. This flux is given by Fick's law of diffusion: Page 9 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College F dm 4R 2 D dt R where D is the molecular diffusion coefficient for water vapour in air and is vapour density such that e = RvT. Integrating this equation from the surface of the drop out to infinity leads to: dm 4rD ( r ) dt where is simply the environmental vapour density. It can be seen that if r the drop will grow, whereas if r it will shrink. The problem now is to establish r' which depends on droplet size, chemical composition and temperature. In turn the temperature is determined in part by the exchange of latent heat which must therefore be considered. In the case of condensation the heat release keeps the droplet temperature higher than that of the environment. The diffusion of heat away from the droplet can be treated in a way analogous to the diffusion L dm 4rK (Tr T ) dt of water vapour to the droplet, giving the equation: where L is the latent heat of vaporisation of water and K is the coefficient of thermal conductivity of the air. Note: the transfer of heat is from the droplet to the environment. The drop temperature and the vapour density at its surface are related by the equation of state for water vapour combined with the formula for equilibrium vapour pressure over a curved solute surface: r e s' (r ) a b e s (Tr ) 1 Rv Tr r r 3 Rv Tr No analytical solution exists for this equation, but it may be solved numerically for temperature and vapour pressure at the droplet surface to obtain the rate of growth by condensation (Mason, 1971). r dr S 1 .......... .......... .......... .......... .......... .....( 4) dt FK FD Here S is the relative humidity (S - 1 is the supersaturation). Notice as r increases, the slower the growth rate becomes. FK represents a thermodynamic term associated with heat conduction, whilst FD is a water vapour diffusion term. The greater the thermal conductivity (K) of the air, the smaller is FK and therefore the faster the rate of droplet growth. Similarly, the greater the water vapour diffusion coefficient (D), the smaller FD and the faster the rate of droplet growth. Including the Kelvin and solute effects: Page 10 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics dr r dt a b r r3 .......... .......... .......... .......... .......... ...( 5) FK FD ( S 1) Again, it is not possible to integrate this equation analytically to obtain drop size as a function of time, though numerical methods can be employed. Table 3 gives growth rate results for droplets growing on nuclei of sodium chloride at 0.05% supersaturation, p = 900hPa, T = 2730K. Table 3 Time (secs) for droplets of initial radius 0.75m to grow by condensation to stated radius (Mason, 1971). Nuclear mass (kg NaCl) 10-16 10-15 10-14 1 2.4 0.15 0.013 2 130 7 0.61 4 1000 320 62 10 2700 1800 870 20 8500 7400 5900 30 17500 16000 14500 50 44500 43500 41500 Radius (m) 4.1 Narrowing of the droplet spectrum To summarise: a droplet forming on an initially large condensation nucleus grows faster but, after a certain radius (about 10m in Table 3), growth rates become similar since the a/r and b/r3 terms are no longer important. The expression (4) for droplet growth rdr/dt can be integrated: r (t ) r02 2t Consider two different sized droplets r2 and r1. r2 (t ) r1 (t ) r2 (0) 2 r1 (0) 2 .......... .......... .......... ........( 6) r2 (t ) r1 (t ) Page 11 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College Thus for two droplets growing in the same conditions of temperature and humidity, the difference between their radii decreases,i.e. the spectrum of droplet sizes becomes narrower (but see Section 5.4). Of course equation (6) can also describe the rate at which a droplet shrinks in an environment of relative subsaturation. Using (6) with expressions for terminal velocity of droplets (which increases with r2 for drops r < 50m), it can be shown that the distance of fall for complete evaporation increases with r 4. Some results are presented in Table 4. Table 4 Distance fallen by drop before evaporating. Assumptions: isothermal atmosphere with T=2800K and RH=80% (Supersaturation, S=0.8). (From Rogers & Yau, 1989). Initial radius (m) Distance fallen 1 2 m 3 170 m 10 2.1 cm 30 1.69 m 100 208 m 150 1.05 km Though results vary with RH, and the expression for terminal velocity used is not accurate for the smaller droplets, these figures show that a droplet has to achieve a radius of order 100m in order to survive the fall from cloud base to ground. Accordingly 100m (0.1mm) is often taken as the dividing line between cloud droplets and precipitation particles - the drizzle drop. 4.2 Growth by condensation of droplet populations Equation (5) describes the growth of a single droplet for given values of temperature and humidity. In order to extend this to the growth of a population of droplets of varying sizes it is necessary to recognise that temperature and humidity will change with time and to consider the water vapour budget of the cloud volume. The rate of change of relative humidity can be written: d dS dz Q1 Q2 dT dt dt where dz/dt is the vertical wind speed and d/dt is the rate of condensation. Q1 represents the increase in saturation through adiabatic cooling on ascent and Q2 represents the decrease in supersaturation due to condensation onto cloud droplets. Together with Page 12 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics equation 4 this equation can be used to examine the evolution of a droplet spectrum, having specified an updraught velocity and an initial droplet distribution (e.g. Mordy, 1959). Typically, droplets growing on the smallest nuclei grow initially but fail to reach the critical radius (r*) and shrink back as haze droplets under the solute effect as supersaturation decreases. The other droplets continue to grow, and the spread in size becomes narrower with time, a property of the growth equation demonstrated earlier. Measurements taken with developing cumulus clouds do indicate narrow drop size distributions centred on a radius in the range 5 to 10m, the average radius increasing with height above cloud base in accordance with the theory of growth by condensation. However, it is observed that higher in the cloud, and in the later stages of development, the spectrum of drop sizes is broader than that predicted by the formulae for growth by condensation - clearly other processes are at work (see Section 5.4). 5. Growth by collision-coalescence 5.1 Collision efficiency: theory and laboratory experiments The series of Figures (7a-d) summarises: (7a) the nature of collision efficiencies between droplets; (7b) theoretical and experiment data on collision efficiencies; drops must achieve a size > ~ 20m before growth by collision can begin; (7c) later studies, which indicate the possible influence of wake capture or ‘micro-turbulence’ in enhancing collision efficiencies as r/R approaches 1. Fig. 7d shows actual collisions between large droplets in the laboratory (Woods, 1965; Mason, 1971). Fig.7a Fig.7b Collision efficiency, E R r • At grazing incidence: E y o2 (R r)2 yo • Collection efficiency=collision efficiency x coalescence efficiency (often put=1) • Collision efficiency, E, for various collector drop sizes as a function of r/R. • Theory & experimental data. E 0 for r<<R and r/R ~1 5.1.1 Collision efficiency and drop size spread • One of the main goals of cloud physicists has been to explain how raindrops could be created in as short a time as 20 minutes from nucleation, a figure often quoted as the observed period between initial Page 13 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College formation of a cumulus cloud and production of rain. As seen already, growth by condensation proceeds quickly at first but slows down with increasing droplet size and it is accepted that there is no way that precipitation may be produced by the condensation process alone in the lifetime of a cloud. Growth by collision-coalescence requires not only the presence of cloud droplets of radius > ~20m, but also a sufficiently broad spectrum of sizes. As seen, growth by condensation, theoretically anyway, makes the spectrum narrower with time. So even if particles grew to 20m by condensation, according to the simple theory there would not be enough • Laboratory Cloud Physics • Computed collision efficiencies, E, for pairs of drops as a function of r/R • A ‘streak’ photograph of the coalescence of two drops, r=62m. • Curves labelled by R • Fall speed, hence drop size, from ‘strobe’ illumination. • Notice ‘wake’ capture effects at r/R~1 Coalescence event • (Woods & Mason 1965) Fig.7c Fig.7d of a spread of sizes for collisions to occur effectively. Various suggestions have been proposed to explain this discrepancy, many involving the entrainment of dry air from the cloud top (Telford and Chai, 1980). See Sections 5.2 and 5.4. 5.2 Growth by continuous collision Consider a droplet radius, R, falling at terminal velocity, u(R) through a population of smaller drops radius, r, terminal velocity u(r). In unit time the average number of droplets collected is: (R+r)2 {u(R) – u(r)} n(r)E(R,r) dr where n(r)dr is the number of droplets per unit volume with radii in the range r to r + dr, E is the collection efficiency (i e. collision efficiency x coalescence efficiency). Integrating this expression to obtain overall increase in droplet volume and then manipulating to obtain dr/dz and dr/dt: Page 14 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics dR EM u ( R) dz 4 L U u ( R) Then assuming droplets radius, r, are sufficiently small, that u(r) = 0 and R + r =R, and that the updraught, U, is negligibly small: dR EM dz 4 L and dR EM u ( R) dt 4 L Here M is the cloud liquid water content (i.e. the mass of cloud water held in drops having radius in the range 0 to R) and E is an average collection efficiency. These are known as the continuous collision equations. Fig. 8 shows calculated growth rates and trajectories of drops, R. While predicting realistic relationships between strength of updraught, cloud height and precipitation size, the results suggest an unrealistically long time to produce precipitation ( >1 hour). Fig.5 Fig. 8 • Calculations of time of droplet growth to precipitation for two updraught speeds. • Assume: collector 20m, droplets 10 m and lwc of 1gm-3 Fig. 8 5.3 Stochastic growth The continuous collision models take no account of statistical fluctuations in droplet concentration, dealing in average values only. A related limitation of earlier analyses was the distinction made between the collector drop and the collected drops. A different approach is to consider the evolution of the droplet spectrum as a whole, rather than classifying a subset of it, as collectors from the start. A coagulation coefficient, K(R, r) may be defined, representing the likelihood that a drop of radius R will overtake, and collide with, a droplet radius r, provided that they are present in unit concentration. Letting V and v Page 15 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College denote volumes corresponding to drop radii R and r, then let H(V,v) represent the probability that a drop of volume V will collect a drop of volume v: 1 1 3V 3 3v 3 H (V , v) K , 4 4 The evolution of the droplet concentration of drops with volume v is determined by two processes - first of all a sink due to coalescences involving drops of volume v to v+dv, and secondly a source due to coalescences between pairs of droplets whose volumes sum to v. So the rate of change of drop concentration in size interval dv is given by: v 0 0 1 dv H ( , n)n( )du n(v)dv H (V , v)n(V )dV .......... .......... ....( 7) 2 where = v - u, in other words and u are two volumes whose combined volume is v. The factor ½ is to correct for the fact that the integral counts each capture combination twice. This stochastic coalescence equation was actually first derived by Smoluchowski in 1916! Fig. 9 shows the result of a modelling experiment based on equation 7, starting from an assumed droplet size distribution at t = 0. Initiation of rain in non-freezing clouds 30 • Example of the computed development of a droplet spectrum by stochastic coalescence. Fig.9 Page 16 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics 5.4 Condensation during coalescence Continued growth by condensation enhances growth by coalescence. Despite the fact that growth by condensation decreases the difference in size between drops, the relative fall speeds do not alter, since the terminal velocity of droplets for r < 40m is proportional to r2 and it has been shown that the difference between the squares of the radii remains constant (equation (6)). Condensation increases the average drop size so average collection efficiency increases, and this effect serves to enhance growth by coalescence (Fig. 10). Effect of condensation on growth by coalescence • The same droplet spectrum is shown evolving a) without, Fig. 10and b) with, an allowance for condensation 5.5 Cloud droplet spectra – the reality Recent studies, while not invalidating the described stochastic and entrainment/mixing etc. processes, do indicate that the development of the required spread in drop sizes is not fully accounted for by such models. The growth rates observed in the atmosphere are greater; indeed the theoretical narrowing of the droplet size spread with time is not observed either. Evidence points to very small-scale vortices enhancing droplet velocities (and hence droplet collision rates) beyond those that could be expected in still air. 6. Ice processes Current research on mixed-phase clouds (Wilson & Ballard, 1999) points to the crucial importance of the ice phase in precipitation development, as summarised in Section 1.1. It must be emphasised that the potential ice nuclei are far fewer than potential cloud condensation nuclei - perhaps one particle in 10 million (1 in 107) qualifies as an ice nucleus at - 200C. Fletcher (1962) estimated Page 17 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College that active ice nuclei are typically present in concentrations of about 1 per litre at - 200C, the number active increasing by a factor of 10 for each additional 40C cooling. That the number of ice precipitation particles can exceed these concentrations may be put down to the various ice multiplication processes discussed in Section 1.1. 6.1 Deposition The equation for growth by deposition of vapour (Bergeron-Findeisen process, Fig.11) is similar to that derived for growth by condensation, being given by: dm 4CD( r ) dt where C replaces r in the condensation equation, and is a function of size and shape of the particle. For a sphere, C = r and the equation is the same as that for growth by condensation onto a water droplet. For a circular disk of radius r, which can be used to approximate plate-type crystals, C = 2r/, whereas ice needles may be approximated by the formula for a prolate spheroid. The deposition growth equation may be derived in a similar way as that for condensation where r is replaced by C, es(T) is replaced by ei(T) and L by Ls (latent heat of sublimation). The equivalent curvature effects are not well understood, and the equation takes no account of the fact that vapour molecules cannot deposit onto ice in a haphazard way - they join in an orderly fashion, molecule by molecule, to maintain the crystal pattern. In addition, the complex shapes of ice crystals mean that the vapour pressure exerted may vary across the ice surface. Experiments suggest that at temperatures between 0 and - 100C small crystals grow about half as slowly as predicted by the depositional growth equation. However, the most important difference between growth by condensation and growth by deposition is the fact that (Si - 1) in the atmosphere is much greater than (S - 1) where water droplets are present, by perhaps 2 orders of magnitude, making growth by deposition a much more efficient mechanism for precipitation production than growth by condensation (Fig.11). Page 18 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics Direct observation of the Bergeron Findeisen process Ice crystal Evaporating supercooled water droplets Fig.11 6.2 Ice nucleation mechanisms Ice may form directly from the vapour phase on suitable deposition nuclei. Three modes of activation are recognised for freezing nuclei (Fig. 12): - some serve first as centres for condensation, then as freezing nuclei. - some promote freezing on contact with supercooled water drops - others cause freezing after becoming embedded in a droplet. A particle may nucleate in different ways - a function of history and ambient conditions; recent work by MRF suggests contact is the Ice nucleation mechanisms Hetergeneous deposition Condensation followed by freezing Contact Immersion Fig.12 Page 19 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College important mode. The relative importance of the different freezing modes has not been established. When in a supersaturated environment it is difficult to distinguish between deposition and freezing nucleation when ice nucleates on an insoluble surface. Even when below water saturation nucleation need not imply deposition because nuclei may contain soluble components. 6.3 Accretion and aggregation Formulae similar to those describing coalescence may be applied to the accretion and aggregation processes, but with less accuracy since the fall speeds of the particles are difficult to specify, depending on their composition (e.g. graupel, snow flakes, clear ice). These different forms are in turn dependent on conditions of temperature, humidity and cloud liquid/ice water content, so the problem is complex (Figs. 13 and 14). Mass and size of different forms of ice crystal Characteristic forms of ice crystals at various temperatures • Ice crystals of differing shapes, growing on a filament in a diffusion cloud chamber with controlled temperature gradient • Mason B J, 1962 • The mass/size relationship is expressed by empirical formulae of the form: m=aDb D: the major linear dimension of the crystal. • Values of a and b for typical crystal forms are, for D in cm, m in gm: • Crystal type a b Graupel 6.5x10-2 3 Thin hexagonal plate 1.9x10-2 3 Stellar crystal 9.4x10-4 2 Planar dendrite 3.8x10-4 2 Needle 2.9x10-5 1 Fig. 13 Fig. 14 6.4 Ice phase versus coalescence An approximate idea of the difference between the rate of precipitation initiation by the ice-crystal process and by coalescence can be gained by comparing the early growth history of an ice crystal with that of a large cloud drop (Rogers & Yau, 1989) (Fig. 15). The ice crystal grows relatively quickly by diffusion, surpassing the initially more massive drop at 75 seconds. After 7 minutes the drop’s collection efficiency is then no longer small and it grows faster than the Page 20 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics crystal. The drop reaches ice crystal mass at 30 minutes (r=160m – a drizzle drop). It has been seen in 6.1 that because of the reduced numbers of ice nuclei compared with droplet nuclei, each ice particle can grow to a large size and achieve a significant vertical velocity, providing a substantial downward flux of moisture. Water drops, however, must spend time coalescing into larger drops before they can remove significant amounts 20 mins 10 mins Ice crystal mass catches up after 7 mins • Times required for an ice crystal and a water drop to grow to a given mass. • Top scale: drop radius. Dashed curves: rate of fractional increase of mass - scale on right of water from the cloud. Fig.15 6.4.1 Mixed phase clouds- and aircraft icing Cloud with tops between 0oC and –4oC are likely to consist entirely of supercooled water drops, leading to aircraft icing problems (Forecasters’ Reference Book, Chapter 2.9.1). With cloud tops of –10oC there is a 50% probability of ice in the cloud; at –20oC the probability increases to 95%. 6.5 Frontal cloud physics Frontal cloud consists predominantly of water in its ice phase. Fig. 16 shows the increasing fraction of ice as temperatures in frontal cloud fall below about –5oC. Page 21 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College Fig.16 Proportion of liquid water Fraction of liquid water in frontal cloud More liquid More ice Temperature (0C) About 1/3 of ice crystals grow by deposition of vapour, about 2/3 by aggregation. Size distributions of ice crystals, constructed from particle image data, obtained during studies by the Meteorological Research Flight, are bimodal, suggesting aggregation of ice crystals as they fall through a cloud. 7. Drop size distributions For steady rain, intensity R, at continental mid-latitudes, the MarshallPalmer (1948) distribution is a reasonable approximation (except a very small drop sizes – Fig.17): N ( D ) N 0 e D where the slope factor depends on rainfall rate : ( R) 41R 0.21 Intercept is given by : N 0 0.08 cm 4 size distributions Where N(D)dD is no. ofMeasured drops sizedrop D to D+dD per unit vol; R is in mm -1 and M-P exponential curves hr 4 10 Break-up may account for the negative-exponential form, raindrops NDlimited in size to D ~ 3mm. being -3 -1 m mm 102 1 5m m m 25 m m hr -1 m hr -1 -1 hr 100 Page 22 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc 0 Fig.17 1 2 D mm 3 4 5 Cloud physics 8. Cloud seeding The introduction to this Note suggested that an early impetus for cloud physics studies was the desire to unravel the possible mechanisms behind seeding clouds with various materials to encourage precipitation. Scientist are much more cautious about their claims now (‘would the cloud have precipitated anyway?’), although there are well documented experiments claiming to confirm that, under certain circumstances, seeding has been effective. Two seeding agents are: silver iodide (AgI), which heterogeneously nucleates ice at a temperature as high as –4oC, and ‘dry ice’ (solid CO2), which has an equilibrium sublimation temperture of -78 oC. The basic premis it that the precipitation-forming processes already present (collision/coalescence, ice processes from natural nuclei) are inefficient. Both seeding agents will vastly increase the numbers of ice crystals in the cloud, which will grow at the expense of the water droplets by the Bergeron-Findeisen mechanism (Section 6, Fig.11). The number of ice particles nucleated will be substantially less than the number of water droplets initially in the cloud, so each ice particle can grow to a relatively large size. The ice particles, being larger than the liquid droplets, fall faster than the droplets they replaced, and can fall to the melting layer to form rain. 9. Lightning – cloud charging mechanisms Mechanisms: Inductive – ‘charge stripping’ is important. Disintegration of large water drops or ice crystals also occurs (Fig. 18). Deformation & disintegration of raindrops • For drops > 6mm diameter aerodynamic pressures exceed surface tension forces. Small scale turbulence encourages instability and breakup. • A large falling drop flattens and develops a depression in its base before breaking up. 2cm • Drop blows up to form expanding ‘bag’ supported by toroidal liquid ring that later breaks into drops • Inductive charge generation is likely in this process. Fig.18 Page 23 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College Thermo-electric effect Supercooled water drop freezes on impact with hailstone. Charge generated across ice-shell + ve carried upwards on splinters • H+ (OH)- + C ice H+,(OH)- Hailstone -ve charge communicated to hailstone W - - 20oC --------- 0oC H+ ,(OH)- More dissociation at W. Diffusion of mobile H+ down concentration gradient: - cold end + ve ---- Ice shell Drop trajectory Supercooled drop Fig.19 Non-inductive – a large number of mechanisms possible, particularly those involving collisions between graupel and ice particles (Fig 19). 9.1 Cloud charge distribution The classic ‘tripole’ charge distribution is apparent in the updraught region (Fig.20). There is basically an upper positive pocket, with the main negative beneath (at about –20 to –25oC), with a positive pocket beneath (at ~ 0oC). A similar but more complex distribution is associated Charge distribution in a thundercloud Fig. 18 -------------------------+++++++ ++++++++++++++++ ++++++++++++++++ ++++ -------- --------- ---------- ------------+++++ +++++ +++++ +++++ Page 24 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics with the downdraught. Fig. 20 9.2 Lightning and cloud microphysics – a summary Lightning and microphysics are closely linked More observational data are required Models must include greater complexity: - more charging mechanisms - improved interactions between ‘processes’ - couplings developed with other models. 10. References and further reading Berry, E et al. 1974 An analysis of cloud droplet growth by collection. J Atmos.Sci, 31, 1814-24 & 1825-31 Blyth, A & Latham J 1998 Comments on glaciation papers by Hobbs et al Q.J.R.Met.S. 124, 1007-8 (*See reply by Hobbs & Rangno) Choularton, T et al 1998 A study of the effects of cloud processing of aerosol on the microphysics of clouds. Q.J.R.Met.S. 124, 1377-1389 Fletcher, N 1962 The physics of rainclouds, Cambridge University Press. Hobbs, P 1993 Aerosol-cloud-climate interactions Academic Press (Ed. P V Hobbs) Hobbs, P et al 1985 Particles in the lower troposphere over the high plains of the United States. J. Clim. Appl. Meteor. 24, 1344 - 1356. Hobbs, P & Rangno* 1998 Reply to: ‘Comments on glaciation papers by Hobbs et al’ Q.J.R.Met.S. 124, 1009-10 IPCC 1995, 2000 and sequels: The science of climate change Page 25 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Met Office College Jonas, P & Mason.B 1982 Entrainment and the droplet spectrum in cumulus clouds. Q.J.R.Met.S. 108, 857-869 Junge, C & McLaren 1971 Relationship of cloud nuclei spectra to aerosol size distribution and composition. J.Atmos. Sci., 28, 382-390 Ludlam, F 1980 Clouds and storms, Pub: Penn State Marshall, J & Palmer,W1948 The distribution of raindrops with size J. Meteor. 5, 165-166 Mason, B 1971 (2nd edition) The physics of clouds. Oxford University Press Mason, B 1975 (2nd edition) Clouds, rain and rainmaking Cambridge University Press. Mason, B 1996 The rapid glaciation of slightly supercooled cumulus clouds. Q.J.R.Met.S. 122, 357-365 Mordy, W 1959 Computations of the growth by condensation of a population of cloud droplets. Tellus, 11, 16-44 Mossop, S 1985 There are two papers on ice particle multiplication in: Q.J.R.Met.S. 111, 113-124 and 183-198 Osborne, S 1996 The processing of aerosols by warm stratocumulus clouds. MRF Internal Note No. 64 Pruppacher, H & Jänicke 1995 Processing of water vapour and aerosols by atmospheric clouds, a global estimate Atmos. Res. 38, 283-295 Püschel, R 1995 Atmospheric aerosols Composition, chemistry and climate of the atmosphere. Int Thomson Publ Inc, 120-175 Rangno, A & Hobbs, P 1994 Ice particle concentrations and precipitation development in small polar continental cumuliform clouds. Q.J.R.Met.S. 120, 573-601 Rogers, R and Yau, M Physics, Slingo, A et al 1989 (3rd edition) A short Course in Cloud Pergamon Press. 1982 Aircraft observations of marine Sc during JASIN. Q.J.R.Met.S. 108, 833-856 Page 26 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc Cloud physics Starr, J 1967 Deposition of particulate matter by hydrometeors Q.J.R.Met.S. 93, (398),516-521 Starr, J 1967 Inertial impaction of particulates on bodies of simple geometry. Ann.Occup.Hyg, 10, 349-361 Taylor,J & McHaffie,A 1994 Measurements of cloud susceptibility J Atmos.Sci, 51, 1298-1306 Telford, J & Chai 1980 A new aspect of condensation theory. Pure & Appl. Geophys. 118, 720 - 742. Toon, O 1995 Modelling relationships between aerosol properties and the direct/indirect effects of aerosols on climate: Aerosol forcing of climate. John Wiley & Sons. Wang, P et al 1978 Effect of electric charges on the scavenging of aerosols by clouds and small droplets J. Atmos. Sci., 35(9), 1735-1743 Wilson, D & Ballard 1999 A microphysically-based precipitation scheme for the UKMO Unified Model, Q. J. R. Met. S. 125, 1607-1636 Woods, J 1965 Wake capture of water drops in air. Q.J.R.Met.S. 91, 585-7 Of general interest: Physical Characteristics of Water: Met. O College Note, Mar-2000 (J Starr) Page 27 of 27 Last Saved Date: 9 March 2016 File: ms-train-college-work-d:\533580417.doc