Precipitation (including rain, snow, and hail) is the primary mechanism for transporting water from the atmosphere back to the Earth's surface. Precipitation amounts and temporal/geographical distribution are affected primarily by atmospheric dynamics; however, precipitation is also influenced by cloud microphysical processes associated with aerosol properties, which are primarily responsible for cloud drop and ice crystal formation. Changes in precipitation regimes and the frequency of extreme weather events (e.g., floods, droughts, severe ice/snow storms, monsoon fluctuations, and hurricanes) are of great importance to life on our planet. Thus, a plausible hypothesis is that by influencing the amounts, chemical composition, and distribution of natural and anthropogenic aerosols, changes in precipitation of significance to local communities may possibly occur. However, quantitative testing of that hypothesis has proved difficult.

Much of the work that was conducted over the years addressed the issues of the effects of aerosols on clouds and was motivated by the desire to understand (so as to predict) natural precipitation formation better as well as to underpin the novel idea of artificial weather modification, following Schaefer's pioneering work on freezing in supercooled water clouds (Schaefer 1946). The fundamental scientific understanding of cloud microphysical processes achieved in those years was summarized sequentially by Fletcher (1962), Mason (1971), and Pruppacher and Klett (1978). Based on many measurements and models, a general consensus emerged that, everything else being equal, the addition of more cloud condensation nuclei (CCN) to a cloud leads to the formation of smaller and more numerous cloud drops. It was also observed that the addition of giant CCN to clouds can lead to the formation of a few larger cloud drops (e.g., Mather 1991) and wider drop size distributions. Furthermore, recent work in shallow orographic clouds has revealed that riming efficiency in polluted clouds is smaller, leading to smaller snow crystals (Borys et al. 2000, 2003). All these observations and related modeling studies suggest as a sound physical hypothesis that, other things being equal, the consequence of particu-late pollution on clouds should be a reduction in precipitation.

Unfortunately, the connection between aerosol concentration and the amount of precipitation on the ground is not yet clear. This is partly because feedbacks between the microphysical, radiative, and dynamical processes must exist and can sometimes lead to enhancement or to suppression of precipitation via atmospheric dynamic rather than cloud microphysical effects of aerosol. Recent examples in support of this proposition on the global scale may be found in the work of Rotstayn and Lohmann (2002), Rotstayn (2007), and Rotstayn et al. (2007). Similarly, warming due to greenhouse gases (GHG) is expected to increase atmospheric water vapor. Global circulation models coupled with mixed-ocean layer models show that increased aerosol concentrations lead to more clouds and thus to reduced solar radiation that reaches the surface. This reduces sensible and latent heat fluxes from the surface and thus reduces precipitation (Liepert et al. 2004). In other words, the effects of aerosols on precipitation could occur through their effects on the radiation in addition to their microphysical effects on clouds. Over the years there have been numerous attempts to shed more light on this connection, but the results vary widely between increases in rainfall, decreases in rain amounts, and no connection at all.

The World Meteorological Organization (WMO) and International Union of Geodesy and Geophysics (IUGG) recognized the importance of this issue and formed a group in 2004 to assess the knowledge and suggest directions for future research. The fi nal report from this group was unable to reach an unambiguous conclusion that a systematic reduction in precipitation is the demonstrated result of particle pollution that enhances CCN levels (Levin and Cotton 2008).

In this chapter, we summarize in three parts some of the main points related to the effects of aerosol-cloud precipitation interactions. We do not intend to give an exhaustive review of the literature, but rather to illustrate the complexities involved in this issue. We begin with a discussion of the issues by separating the potential effects of aerosols on different types of clouds. This is because the effects could be significantly different depending on whether these are convective or orographic, or whether the affected clouds are downwind of urban centers. Using Australia as an example, we explore thereafter the complexities of the evidence and arguments about trends in precipitation and their causal attribution. Finally, we offer some remarks about the manner in which the scientific uncertainty on this issue has been reflected in the community.

The Effects of Aerosols on Clouds

Warm clouds are those that contain no ice. Measurements have shown that an increase in CCN from natural or anthropogenic sources increases cloud drop concentrations and reduces cloud drop size (leading to the first indirect effect or the "Twomey" effect). These ideas have been confirmed by many in-situ measurements following the early work by Squires (1958a, b), Twomey and Squires (1959), and Warner and Twomey (1967), for example by analysis of ship tracks using satellite images (Coakley et al. 1987; Durkee et al. 2000, 2001).

Rosenfeld and Lensky (1998) developed a method of using satellite images to estimate the effective radius of cloud particles near cloud tops as a function of temperature (a surrogate for cloud top altitude). Their analysis suggested that an increase in aerosol optical depth corresponds to slower growth of the cloud drops, because of the increase in their number concentrations and decrease in their effective radius. Slow growth in the warm clouds may lead to elimination or the suppression of precipitation development. Radar echoes from the TRMM satellite in combination with the estimate of the effective radius at cloud top were interpreted as showing that the development of precipitation in these clouds diminishes, although the number of such documented cases has been small and is controversial (Ayers 2005; Rosenfeld et al. 2006).

Airborne and ground measurements of clouds and precipitation in the Amazon region (Andreae et al. 2004) showed that clean continental clouds with relatively low numbers of aerosol particles, behaved similarly to marine clouds; namely, growth by coalescence occurs early and rapidly as the cloud develops. On the other hand, clouds that developed in the smoky atmospheres grew deeper with precipitation particles growing higher up in the clouds. Andreae et al. (2004) argued that the slow growth of the drops led to ice formation higher up in the clouds and to enhanced updrafts attributable to an increase of condensed water and resultant increased release of latent heat. Such clouds sometimes led to hail and lightning formation. These conditions were even more pronounced when the clouds developed just above the fire region (named pyroclouds). In such cases, input of heat and large numbers of smoke particles led to invigoration of the cloud, resulting in taller clouds and larger cloud particles. There is no evidence, however, that these clouds produced more or less rain amounts on the ground.

Aerosol Impact on Rainfall on the Ground Convective Clouds

Warner and Twomey (1967) and Warner (1968) summarized the potential effects of sugarcane smoke on rainfall by looking at multidecadal rainfall records from stations up- and downwind of these prolific anthropogenic aerosol sources. Despite the expectation that there would be a direct correlation between increased pollution and rain suppression, they could not conclusively see one (Warner 1971). In his 1971 paper, Warner stated, "It would be surprising if the microphysics of a cloud played no part in determining its rainfall, but we must await further results if this is to be adequately demonstrated."

Rain enhancement of up to 30% from warm clouds downwind of paper mills in the state of Washington was reported by Hobbs et al. (1970). Analyzing the same case through the use of a one-dimensional numerical model, Hindman et al. (1977) concluded that the emitted giant CCN from the paper mill could not by themselves account for the observed large increase in rainfall, and that the total effects of heat, water vapor, and CCN from the paper mill in combination may be responsible for the increased precipitation.

It is valuable to note that Mather (1991; Mather et al. 1997) reported an increase in radar echo from mixed-phase convective clouds affected by particles emitted from paper mills in South Africa. His observations led to a large field experiment for rain enhancement using hygroscopic particles.

Through the use of MODIS and TRMM satellite data, Lin et al. (2006) analyzed the effects of forest fires on precipitation in the dry season in the Amazon region. They reported an increase in cloud heights and in precipitation with increases in aerosol optical depth. The increased cloud height led to enhanced growth of ice crystals, which culminated in heavier precipitation. However, despite the good correlation between these variables, they could not unequivocally establish causal links between aerosols and the observed changes in cloud height or with precipitation increases. The role of enhanced convection attributable to the heat from the fires and/or from absorption of solar radiation by the smoke itself could not be excluded.

There have been many numerical simulations of single convective clouds, suggesting that increase in CCN concentrations leads to reduced precipitation. However, the input of only a few giant CCN (a few per litter) is enough to enhance precipitation (e.g., Yin et al. 2000; Teller and Levin 2006). Teller and Levin (2006) showed that increasing the giant CCN from 1-10 per cm-3 can compensate for a decrease in precipitation amounts due to increased CCN from 200-400 cm-3. Simulating a 3-D case of Crystal Face storm, in which dust particles were present, Van der Heever and Cotton (2006) demonstrated that giant CCN could affect the rain intensity early in the development of the storm by permitting higher supercooled water contents to be lifted to higher levels and freeze. Several hours later the enhanced CCN simulations produced less precipitation because the cold pools, which were associated with the evaporation of the falling hydrometeors, decoupled themselves from the sea breeze convergence zone. The overall effect is, therefore, not straightforward to allow predictions with certainty.

Effects of Urban Pollution on Rainfall

Extensive studies were conducted to explain the seemingly anomalous behavior of the precipitation around La Porte, Indiana, which is located downwind of Chicago. Local records suggest an upward shift in warm season rainfall, thunderstorms, and hail from the late 1930s to about 1965. The puzzling thing about this case is the fact that the anomaly appeared and then disappeared. In reviewing the observations, Changnon (1980) concluded that the microphysi-cal effects must have played a role but without some dynamic or meteorological influence, this effect could not have occurred.

In the 1970s, a large field experiment (METROMEX) was conducted around St. Louis, Missouri, motivated by the examination of historical data, which revealed summer precipitation increases in the immediate downwind area of the city (Figure 16.1). The records show increases in rainfall (10-17%), moderate rain days (11-23%), heavy rainstorms (80%), thunderstorms (21%), and hail storms (30%) (Changnon et al. 1971). In his summary of METROMEX, Braham (1974) reported that the CCN production from the city was about 104 cm-2 s-1 (i.e., much higher than was present in the surrounding rural areas), accompanied by an increase in cloud drop concentrations and a decrease in drop

1941 1945 1949 1953 1957 1961 1965

Start of 5-yr period

1941 1945 1949 1953 1957 1961 1965

Start of 5-yr period

Figure 16.1 Five-year moving averages and time trend of Centerville (downwind of St. Louis) summer rainfall, 1941-1968 (after Changnon et al. 1971).

Figure 16.1 Five-year moving averages and time trend of Centerville (downwind of St. Louis) summer rainfall, 1941-1968 (after Changnon et al. 1971).

size. However, radar echoes from the large droplets in these clouds usually occurred lower in the atmosphere than their counterparts in the rural surroundings. This seems to contradict our physical understanding of cloud growth; however, Braham concluded that one way to explain the observations was to assume that the urban area also emitted giant CCN, which were not detected by the CCN sampling methods in use, but which could be responsible for the increased precipitation.

More recently, Van der Heever and Cotton (2007) simulated the effects of pollution on precipitation during the passage of a storm in the St. Louis area. The thermodynamic data that they used were from a recent specific day in which ordinary thunderstorms were prevalent, whereas aerosol input data were based on averages obtained during the METROMEX experiment. The simulation was carried out using two main cases: one with the city of St. Louis without pollution and the other with the pollution containing both small CCN and giant CCN. The model results show the expected temporal and spatial distribution of the rain changes as a result of the effects of pollution (Figure 16.2). At the beginning of the storm, polluted clouds produced much heavier precipitation; however, as the storm progressed, the difference between the integrated amounts of rain from the beginning of the storm of the polluted minus the clean case diminished. After 1.5 hour, the integrated rain amount over the whole area was higher in the clean case. This work demonstrates the complexity of the interaction between aerosols and precipitation. Part of the complexity appears attributable to the fact that the initial rain cleans the atmosphere of pollution, thus reducing the effects that pollution will have on further rainfall. In addition, downdrafts produced by the precipitation enhance the development of neighboring clouds, thus increasing the integrated rain amounts over the whole

Figure 16.2 Model results showing accumulated surface precipitation from clean-polluted clouds around the city of St. Louis, Missouri. Solid lines represent pollution suppressing precipitation; Dash lines represent the opposite. Contour interval is 5 mm starting from 1 mm. Note the changes in spatial and temporal distribution of rain (after Van Den Heever and Cotton 2007).

Figure 16.2 Model results showing accumulated surface precipitation from clean-polluted clouds around the city of St. Louis, Missouri. Solid lines represent pollution suppressing precipitation; Dash lines represent the opposite. Contour interval is 5 mm starting from 1 mm. Note the changes in spatial and temporal distribution of rain (after Van Den Heever and Cotton 2007).

area. Van der Heever and Cotton (2007) concluded that the effect of pollution on precipitation in an urban setting depends strongly on the background aerosol concentrations. Adding more pollution to an already polluted atmosphere has very little effect on precipitation amounts. They further indicated that effects of land use (e.g., soil moisture, release of sensible and latent heat fluxes, modification of wind stress and more) play a dominant role in those precipitation anomalies. A key factor, as far as precipitation response is concerned, is whether the secondary convection (forced by cold pools) remains coupled to the urban land-use driven mesoscale circulation or not.

Using four years of observation data obtained from the NASA Earth Observing System (EOS) MODIS, in-situ AERONET, and in-situ EPA PM2.5 data for one mid-latitude city (New York) and one subtropical city (Houston), Jin et al. (2005) analyzed diurnal, weekly, seasonal, and interannual variations of urban aerosols with an emphasis on summer months. Their research reveals that spatial and temporal urban aerosol optical depth varies dynamically as a result of various parallel factors: human activity, land-cover changes, cloud-aerosol interactions, and chemical processes. Diurnal, seasonal, and interannu-al variations of aerosol optical depths were examined and found to be largely affected by weather conditions; however, the optical depth peaks often during the rush hours in the morning and evening. Analysis of monthly mean aerosol optical thickness and rainfall did not show strong relationships between aerosol and rainfall in a climatological sense.

In this analysis, virtually no seasonality was observed for rainfall over Houston and New York City, suggesting that aerosols affect rainfall less than larger-scale processes (e.g., land use, urban heat island effects). Around Houston, the TRMM satellite-based accumulated rainfall data show that the maxima in monthly mean rainfall occurred in October 2000, May 2001, and September 2002. This is consistent with the transition between seasons in this region. In general, New York's rainfall had less month-to-month variation than Houston, with a maximum slightly above 200 mm/month in October 2002. The effective radius for water clouds was lower in New York City than in Houston, suggesting that there were either more aerosols in New York City than in Houston, or thinner clouds. The lack of direct relationship between rainfall and urban aerosol optical thickness implies that urban rainfall anomalies are not fully related to changes in aerosol. This observation is consistent with the earlier conclusions from METROMEX (Ackerman et al. 1978).

As can be seen from the above discussion, in spite of many measurements, there is no conclusive evidence that aerosol pollution from urban regions affects precipitation in a consistent, systematic, or repeatable manner.

Precipitation from Orographic Clouds

Borys et al. (2000, 2003) provide some evidence that pollution can delay precipitation in winter orographic clouds in the Rocky Mountains. Their analysis shows that pollution increases the concentration of CCN, and thus cloud drops, leading to the formation of smaller cloud drops. The reduced drop size leads to less efficient riming and therefore to smaller ice crystals (Figure 16.3), smaller fall velocities, and less snowfall.

Givati and Rosenfeld (2004) analyzed about 100 years of records of oro-graphically forced convective precipitation in regions located downwind of pollution sources and compared them to precipitation in regions unaffected by these sources. In their study, they documented the precipitation trends in the orographic enhancement factor, Ro, which is defined as the ratio between precipitation over the hill with respect to the upwind, lowland precipitation amount (Givati and Rosenfeld 2004). Two geographical areas were chosen for this study: California and Israel. The topography in both regions is similar, although the mountains in Israel are much lower than the Sierra Nevada. Their statistical results for both locations show that downwind of pollution sources, on the upslope of mountains and mountain tops, orographic precipitation is reduced by ~20% and ~7%, respectively. It was hypothesized that this decrease is attributable to an increase in droplet concentrations and a decrease in droplet size. Farther downwind on the lee side of mountains, the amount of precipitation increased by ~14%. Givati and Rosenfeld postulated that this increase is because smaller cloud particles take a longer time to grow, allowing the winds aloft to carry them over the mountain top (see earlier study of similar effects, produced by deliberate over-seeding with ice-producing particles, by Hobbs 1975a, b). However, they hypothesized that the integrated rainfall amount over the whole mountain range was reduced by (an assumed but not substantiated) progressively increased pollution over the years. Subsequent studies show similar decreasing trends in Ro over a few western states in the U.S. (Rosenfeld and Givati 2006; Griffith et al. 2005) and the eastern slopes of the Colorado Rockies (Jirak and Cotton 2006). Givati and Rosenfeld argue that although absolute precipitation amounts and Ro are affected by fluctuations in the atmospheric circulation patterns, such as those associated with the Pacific mrn'm.

Figure 16.3 Light riming of ice crystals in clouds affected by pollution (left) compared to heavier riming in non-polluted clouds (right). (From Borys et al. 2003, by permission of the American Geophysical Union.)

Decadal Oscillation and the Southern Oscillation Index, these cannot explain the observed spatial variational trends in Ro.

Recently, Alpert et al. (2008) re-analyzed the data from Israel using the orographic ratio method, taking the ratio between the stations on the Samaria Mountain and the stations located upwind along the seashore, as well as stations on the mountains in the western Galilee and the seashore near the city of Haifa. Their results show effects opposite to those reported by Givati and Rosenfeld (2004, 2005); namely, the orographic ratio actually increased (see Figure 16.4) over the years. Alpert et al. concluded that at least in Israel, factors other than aerosol pollution dominate the precipitation amount in orographic clouds. They showed that by calculating Ro for all the stations on the mountain against the stations along the coast (and not stations located in and downwind of the urban centers, as was done by Givati and Rosenfeld), there are more cases in which Ro increased rather than decreased over the years (Figure 16.5). It can be clearly seen that the majority of pairs show an increase in orographic ratio between the Samaria Mountains and the seashore, in contrast to the conclusion of Givati and Rosenfeld (2004). Alpert et al (2008) argued further that the orographic ratio is not an appropriate method to estimate the effect of pollution on rainfall. This is because the ratio can decrease not only by decreasing the numerator but also by increasing the denominator. To make things worse, many of the stations upwind of the mountains used by Givati and Rosenfeld (2004, 2005) were located in or downwind of urban pollution sources, where the rain actually increased over the years. Thus, these stations were affected not only by pollution but also by other, probably much more important, factors such as urban effects (e.g., urban heat island, changes in frictional velocity,

y = 0.006x + 1.088 R2 = 0.176 p = 0.007

* ♦ ♦ ♦ ♦«


'___— ---♦


♦______♦♦__________♦ ♦_______

♦ ♦♦

______♦ ♦_______________ .

♦ * *



Figure 16.4 The annual precipitation ratios between the Samaria hills and the central coast clusters, for the period 1952-1998, are plotted along with the best-fit line. The dates on the abscissa represent the winter season (November to April), which is the rainy period in Israel. Note the significant increasing trend of the orographic ratio (r = 0.42, p = 0.007) in contrast to the results of Givati and Rosenfeld (2004). After Alpert et al. (2007).

100.1 - 200 1200.1 - 300 300.1 -400 400.1 -500 | 1500.1 -600 [7 i 600.1 -700 m 700.1 -£00 ■1890,1 -900 H 900.1 -1.000 - i.ioo

* *



* ^


/ *

Figure 16.5 The orographic ratio (Ro) in central Israel between mountain stations and those along the shore and inland. White lines indicate an increase in Ro over the past fifty years and black lines indicate the opposite. The thicker the lines, the larger Ro is. Note that most Ro indicate an increase over the years, in contrast to the report by Givati and Rosenfeld (2004). Figure based on Alpert et al. (2008).

Figure 16.5 The orographic ratio (Ro) in central Israel between mountain stations and those along the shore and inland. White lines indicate an increase in Ro over the past fifty years and black lines indicate the opposite. The thicker the lines, the larger Ro is. Note that most Ro indicate an increase over the years, in contrast to the report by Givati and Rosenfeld (2004). Figure based on Alpert et al. (2008).

land-use change). Such urban effects on increased downwind precipitation have been found by many other investigators in many other locations (e.g., Braham 1974; Landsberg 1981; Goldreich 2003; Goldreich and Manes 1979).

In a more recent publication, another version of the orographic analysis was presented on hypothesized suppression of precipitation attributable to pollution in China (Rosenfeld et al. 2007). This paper relies on data from very few rain gauge stations near two cities and one station on the mountain top. Rosenfeld et al. (2007) reached strong conclusions about the suppression of precipitation by pollution, by making a connection between visibility and rainfall, where visibility is used in the paper as an indicator of pollution levels. However, their results show (Figure 4 in Rosenfeld et al. 2007) that under the same low visibility conditions, the amount of precipitation in the urban area increased (not suppressed) more than over the mountain. Thus, in contrast to the conclusion by the authors, using the orographic ratio Ro, it is impossible to separate increases attributable to the urban effects (including pollution in the urban area) from the factors affecting precipitation over the mountain. This reinforces one of the points made by Alpert et al (2008), and mentioned above: care must be taken in the choice of the stations used for evaluating the orographic ratio. It is clear that inherent variability in the climate system and changes in local population distribution will always yield particular regional (or rain gauge) patterns that, depending on choice of rain gauge, can yield diverse signals. Moreover, simple empirical correlation of surrogate variables based on historical time series cannot take into account the evident non-stationary nature of climate known to be associated with anthropogenic global warming as well as natural decadal variability. New and better statistical methods need to be developed to separate the different factors affecting precipitation.

In summary, although pollution does affect clouds, the hypothesized effects on precipitation are not yet clearly understood. In fact, even the effects of clouds on aerosols (e.g., scavenging of aerosols during the rainy period) have hardly been considered in these correlation studies. It seems quite probable that other factors, such as synoptic-scale processes, urban effects, and mesoscale dynamic factors, dominate the precipitation amounts over the mi-crophysical processes, while long-term trends cannot be divorced from changing weather patterns driven by global warming and other long-term drivers of climate change. We will delve into this complexity by closely looking at what we know about the effects of pollution on precipitation in Australia, as but one possible case study, and then try to refl ect more broadly about the potential effects elsewhere.

Complexity at Regional Level: Australia

Australia provides a useful example of the complexities inherent in seeking evidence of an aerosol-induced impact on precipitation. As a case study, it has singular advantages: it is an isolated continent effectively immune from longrange transport of particle pollution from elsewhere; it is the driest continent, and thus has a long history of scientific interest and effort in atmospheric and cloud physics research relevant to the question at hand; it is relatively unpopulated (approximately 21 million in 2006) comparable with its land area (ca. 7,600,000 km2); and the majority of its population (~15 million, or almost two thirds) inhabit just eight cities (the State and Territory capitals), leading to strong contrasts in air composition between the isolated city air sheds and the vast, clean regions in between. Thus experiments contrasting "clean" and "polluted" conditions, un-confounded by long-range transboundary air pollution, are possible in Australia, which would be impossible in most countries of the northern hemisphere (Bigg and Turvey 1978).


From the 1950s through the 1980s, Twomey and Bigg spent considerable effort trying to understand the characteristics and climatology of CCN, condensation

(or Aitken) nuclei (CN), and ice nuclei (IN) in Australia, and more generally. Of particular note, with respect to CCN, was Twomey's systematic and comprehensive investigation of variations in cloud-active aerosols at the mid-latitude site of Robertson, New South Wales. This site was located in a rural region well over 100 km to the southwest of the mega-metropolitan region covered by Sydney; it was due west of the heavily industrialized steel-making city of Wollongong, due east of the rural township of Bowral, and southeast of the twin towns of Mittagong and Moss Vale. Twomey et al. (1978) showed that air masses arriving at Robertson after passage over the major cities Sydney or Wollongong had average CCN levels that exceeded 800 cm-3, air masses passing over the smaller towns of Mittagong and Moss Vale had concentrations averaging 400-800 cm-3, air from Bowral typically contained 300 cm-3, and air masses from other, relatively unpopulated regions exhibited CCN levels of only ~150 cm-3 (Figure 16.6). Clearly anthropogenic activities enhanced CCN levels severalfold over the natural, continental background levels, with plumes of cloud-active aerosols wandering over the Australian land mass as dictated by location of the urban areas and meteorology.

Bigg and Turvey (1978) reported on the results of an ambitious airborne program from 1974-1977 that mapped the distribution of CN across the whole continent as a function of altitude from the surface to about 6000 m in altitude. Their work reinforced and amplified Twomey's findings: median CN concentration over the continent in the well-mixed and relatively unpolluted boundary layer was very low by world standards at 680 cm-3 (very consistent with Twomey's CCN level of ~150 cm-3 for "clean air," given that CCN are a subset of CN). In contrast, CN plumes from population centers were greatly enhanced in concentration, with Bigg and Turvey (1978) stating:

Particle production by one of the smaller metropolitan areas at 4 x 1019 s 1 was shown to exceed the estimates from all natural sources, indicating that particles resulting from human activities may often dominate the continental aerosol over much of Australia.

Figure 16.6 Wind rose of CCN concentrations for air masses arriving at Robertson, NSW, from 1968-1973. Reproduced from Twomey et al. (1978).

Particle production by one of the smaller metropolitan areas at 4 x 1019 s 1 was shown to exceed the estimates from all natural sources, indicating that particles resulting from human activities may often dominate the continental aerosol over much of Australia.

Figure 16.6 Wind rose of CCN concentrations for air masses arriving at Robertson, NSW, from 1968-1973. Reproduced from Twomey et al. (1978).

Bigg extended his measurements to evaluate total particle emission flux over Australia by developing a methodology in which an aircraft made continuous CN measurements between the surface and the boundary layer inversion at fixed distances downwind of the source. Convolved with wind speed measurements also taken by the aircraft as a function of altitude, the downwind plume CN cross section yielded an instantaneous transport rate through the sampled slice of plume, or an instantaneous "emission flux" measured at the given downwind distance. This methodology was then applied to the isolated Mt. Isa copper and lead smelter complex in tropical western Queensland as the experimental source. The repeated airborne flux measurements confirmed yet again that anthropogenic aerosol plumes exhibited greatly enhanced CN and CCN concentrations over natural, continental levels, and also traveled great distances (> 600 km). Ayers et al. (1979) and Bigg et al. (1978) provide examples of plume cross sections measured, plus derived fl ux estimates that decrease out to 200 km from Mt. Isa but increase thereafter as gas-to-particle conversion of the SO2 in the smelter plume created new particles by homogeneous nucleation. Subsequently, Carras and Williams (1981) tracked the Mt. Isa plume to an astonishing 1800 km downwind, at which point it exited the Australian continent over the Indian Ocean. The importance of anthropogenic activities as a source of CCN is underscored by the estimate of Ayers et al. (1979) that Mt. Isa alone had a CCN emission flux equivalent to ~0.1% of the total, natural, global emission flux for CCN active at 0.5% supersaturation.

The same methodology was applied to a program of particle emission flux measurements from a representative number of urban centers across the country. Figure 16.7 shows CN flux as a function of population downwind of a range of towns and cities (Ayers et al. 1982). A theoretical analysis by Manton and Ayers (1982) confirmed the physical plausibility of the essentially linear

1,000 10,000 100,000 1,000,000 City population

Figure 16.7 Condensation nuclei flux as a function of population downwind of a number of Australian towns and cities (Ayers et al. 1982).

■ Whyalla o Kalgoorlie o Mt. Isa a Gladstone o Geraldton

□ Carnarvon a Cloncurry o Broome

1,000 10,000 100,000 1,000,000 City population


Figure 16.7 Condensation nuclei flux as a function of population downwind of a number of Australian towns and cities (Ayers et al. 1982).

Figure 16.8 Model of the distribution of primary condensation nuclei concentration at 500 m over SE Australia on October 21, 1998, at 16.00 local time. Source function used was the relationship shown in Figure 16.6 (Ayers et al. 1982) convolved with a gridded population density database available at 1 km resolution. The chemical transport model used for this simulation was TAPM run at 12 km resolution.

Figure 16.8 Model of the distribution of primary condensation nuclei concentration at 500 m over SE Australia on October 21, 1998, at 16.00 local time. Source function used was the relationship shown in Figure 16.6 (Ayers et al. 1982) convolved with a gridded population density database available at 1 km resolution. The chemical transport model used for this simulation was TAPM run at 12 km resolution.

relationship over more than three orders of magnitude evident from Figure 16.7: the flux-population relationship is 8 x 1013 particles per person per second. Use of this figure plus gridded population density, as a surrogate explanatory variable for primary CN emissions from human activities in Australia, enables modeling of emissions from towns of all sizes in any part of the country, as illustrated in Figure 16.8. Extrapolation to the whole continent based on total population led immediately to the conclusion that total anthropogenic particle production over Australia, at ~1021 s-1, exceeded the natural particle emission rate by more than an order of magnitude (Bigg and Turvey 1978).

Aerosol and Cloud Properties

The role of variability in CN and CCN concentrations as a determinant of variability in cloud microphysical structure at the condensation level has been studied in a number of Australian experiments over both the continent and ocean. Here we focus only on continental examples; for completeness, we note that additional work paralleling Bigg's marine CN studies addressed the marine environment upwind of Australia. One marine example is the work of Boers et al. (1994, 1998) and Boers and Krummel (1998), which couple data on long-term, systematic seasonal variations in CCN measured at Cape Grim (determined by the seasonal cycle in dimethylsulfide emissions from the Southern Ocean) with airborne measurements in the Southern Ocean Cloud Experiment (SOCEX) on marine stratocumulus properties upwind of Cape Grim. Figure 16.9 shows the remarkable agreement achieved between average

0123456789 10 11 12 Month

Figure 16.9 Measured (SOCEX) summer (January) versus winter (June) average droplet concentrations and effective radii near cloud base in marine stratocumulus clouds upwind of Cape Grim, Tasmania, compared with seasonal cycles modeled using an explicit cloud model and the observed monthly mean CCN concentrations measured at Cape Grim (Boers et al. 1994, 1998).

summer and winter cloud droplet concentrations just above cloud base and droplet concentrations predicted by a cloud model with explicit CCN-droplet size dependence. Here, we consider three continental examples in detail.

Effects of Smoke from Sugar Cane Fires

Warner and Twomey (1967) presented subcloud CCN and in-cloud droplet concentration data from just above the condensation level taken in airborne surveys up- and downwind of massive sugar cane fires in NE Queensland (removal of foliage by burning prior to harvest was at that time a normal practice). Their data has been replotted in Figure 16.10, demonstrating a clear

200 "

0 200 400 600 800 1000 1200

Drop concentration computed from CCN (cm-3)

Figure 16.10 Relationship measured between subcloud CCN concentrations and cloud droplet concentration near cloud base upwind and downwind of sugar cane fires in NE Queensland (Warner and Twomey 1967).

-•- Modeled (from CCN) — Measured (SOCEX)

relationship between CCN numbers below and cloud droplet concentrations above the cloud base.

Effects of Mt. Isa Copper Smelter Plume

Comparable results were obtained in studies conducted 56 km downwind of Mt. Isa in which cloud droplet size distributions were determined just above the condensation level in fair-weather cumulus clouds growing in air masses inside and outside the sulfate aerosol plume from the Mt. Isa smelter complex. Figure 16.11, redrawn from Ayers (1981), depicts cloud droplet spectra measured in two sets of clouds on March 5 and 6, 1977. Clearly evident is an elevation in cloud droplet number comparable to that noted by Warner and Twomey in smoke-affected clouds downwind of cane fires, plus the concomitant reduction in cloud droplet size that must accompany the concentration increase, assuming that everything else (e.g., liquid water content, atmospheric temperature and humidity profiles, and synoptic conditions) is the same.

Effects of the Adelaide Plume

Application by Rosenfeld (2000) of the Rosenfeld and Lensky (1998) algorithms to a remotely sensed cloud field recorded over Adelaide on August 12, 1997, showed unmistakable evidence of "pollution tracks" in clouds downwind of Adelaide and nearby rural cities. These pollution tracks showed up as broadening regions of high-concentration/low-effective radius droplets in cloud. This pattern, observed using remote-sensing technology, confirmed the evolution of spreading CN plumes demonstrated downwind of Adelaide and other Australian cities by Bigg and Turvey (1978), Ayers et al. (1979, 1982), and Manton and Ayers (1982). The explicit connection between the city as a



JD 1

539 cm-3

■ Inside plume - Outside plume

Lr h

180 cm-3

20 40 60 80 100 120 Diameter (|jm)


691 cm-3

156 cm-3

— Outside plume

20 40 60 80 100 120 Diameter (jm)

Figure 16.11 Droplet distributions measured close to the cloud base in fair weather cumulus clouds near Mt. Isa growing in air inside or outside the plume from the Mt. Isa copper smelter (Ayers 1981). Top: March 5, 1977; bottom: March 6, 1977.

prolific aerosol source and a resultant increase in cloud droplet numbers or decrease in cloud droplet size provided a modern reconfirmation of the earlier works of Warner, Twomey, and Ayers in NE Queensland and at Mt. Isa.

Precipitation Suppression

The previous sections provide compelling evidence that even over a relatively unpopulated continent, such as Australia, anthropogenic aerosols dominate the aerosol number concentration, and that where these aerosols are mixed up into cloud, they affect cloud microphysical properties in the anticipated ways: increasing cloud droplet number concentration and decreasing droplet effective radius. Although this clearly has the potential to affect atmospheric radiative transfer via an effect on cloud albedo, as foreshadowed by Twomey (1977), it is not the so-called "first indirect effect" (IPCC 2001) that is the focus here, but rather the subsequent effect, if any, on formation of precipitation.

It has been known for some time that a broad droplet distribution with a significant population of cloud droplets (radii > 15 ^m) is needed for precipitation to occur efficiently in warm clouds (Fletcher 1962), as has the hypothesis that polluted clouds having small droplets and narrow droplet size distribution might be less efficient in production of rain. However, only recently has "rainfall suppression by pollution" been claimed to have been demonstrated (Rosenfeld 2000). From experience, we know that no matter how clear the effects of aerosols are on cloud microphysical properties, it has proved extremely difficult to test the subsequent hypothesis that microphysical changes significantly alter precipitation. This comment stems from the 50-year history of an inability by the scientific community to demonstrate unequivocally that well-designed, intentional weather modification experiments based on aerosol dispersal in cloud can demonstrate altered precipitation patterns at scales above that of single clouds. It is worth emphasizing the need for caution on this issue by restating a key introductory sentence from the U.S. National Academies report (BASC 2003): "The committee concludes that there is still no convincing scientific proof of the efficacy of intentional weather modification efforts."

Effects of Smoke from Sugar Cane Fires

The demonstration by Warner and Twomey (1967) that biomass burning aerosols from cane fires dramatically altered downwind cloud properties provided Warner with a natural laboratory in which to seek evidence of concomitant reductions in rainfall associated with aerosol pollution. The experimental design was relatively straightforward. Warner (1968) presented an analysis of 60 years of rainfall records for stations up- and downwind of the cane-growing regions of NE Queensland, and sought a decrease in downwind rainfall with time that mirrored the history of expanding cane production. Warner produced the cautious conclusion that such a signal appeared to be evident in the data, but noted that "the possibility that other factors caused the particular climatic changes observed cannot be eliminated."

Based on this promising result, Warner (1971) went on to conduct a more comprehensive analysis, which was presented at the 1971 International Conference on Weather Modification in Canberra. However, this more complete analysis led him to revise the earlier conclusion: "What has been said above may explain why it was not possible to detect an association between increasing cane production and rainfall. Nevertheless, we must conclude that the present study gives no support to the idea that association found between cane fires and rainfall at Bundaberg was due to inhibition of the coalescence process by a reduction in average cloud droplet size." A null result.

Effects of the Adelaide Plume

In contrast to Warner's carefully argued null result, Rosenfeld (2000) claimed positive evidence for long-term suppression of rain and snow formation through anthropogenic aerosols downwind of Adelaide. A careful reading of Rosenfeld's paper, however, suggests a need for caution. His conclusion is not based on a representative sample, since it is based on the detailed analysis of only a single satellite image, which is an impossibly small "instantaneous" sample from which to reach such a substantial conclusion concerning long-term spatial and temporal rainfall patterns in SE Australia. Moreover, there was an absence of supporting evidence to validate any of the (a) remotely sensed cloud properties, (b) pattern of aerosol pollutants across the remotely sensed cloud fields, (c) spatial distribution of liquid water content in the clouds, and (d) pattern of rainfall at the ground. Each of these factors was considered in detail by Ayers (2005), who concluded: "In the case of rainfall suppression over Australia presented by Rosenfeld (2000), the analysis presented above identifies major doubts about the conclusions reached; indeed the conclusions in that work are invalid based on the evidence available. This is the same position reached by Warner (1971) almost two decades earlier." A rebuttal of this conclusion was subsequently published by Rosenfeld et al. (2006). However the rebuttal fails to address either the problems of detail or the two fundamental flaws evident in Rosenfeld (2000): neither (a) the experimental design used in that work nor (b) the specific day chosen for study is capable of providing a scientifically valid test of the rainfall suppression hypothesis.

Precipitation in Australia


Rainfall trends over the last century across Australia have been analyzed by a number of people (e.g., Suppiah and Hennessy 1998; Hennessy et al. 1999; Smith 2004). For all-Australian rainfall, Smith (2004) identified a positive long-term trend over the century, though differences in sign of the trend and with respect to season, along with high variability, emerge when the data are broken down to regional scale (Suppiah and Hennessy 1998; Hennessy et al. 1999; Smith 2004). Significantly, a generalized Australia-wide long-term suppression of rainfall caused by increasing population and increasing pollution emissions over the decades is not supported by these data: it would be illogical to argue rainfall suppression on average when long-term rainfall on average has increased. This is also the case when the southeast region discussed by Rosenfeld (2000) is considered. Figure 16.12 plots the average rainfall time series for SE Australia, which reveals a positive trend up until 1995, followed by a sharp reduction thereafter. This stands in sharp contrast to Rosenfeld's qualitative claim that rainfall decreased over decades in this region since pollutant emissions have grown slowly with population. The more so, since the decline, post-1996, corresponds with the implementation of a National Environmental Protection Measure for Air (the Air NEPM) in 1998, which resulted in a stabilization of urban air quality in the decade thereafter, during which rainfall sharply declined. Evidently the recent rainfall decline cannot be ascribed a priori to a concomitant increase in level of pollution. It is equally evident that simple correlations between rainfall and assumed or implied levels of "pollution" do not provide a particularly robust test of the rainfall suppression hypothesis. This is because the inherent variability in the climate system will always yield particular regional (or rain gauge) patterns that, depending on choice of rain gauge, can be interpreted in a variety of ways (see the discussion above relating to the effects of pollution on orographic precipitation in Israel). We conclude that in the absence of comprehensive time series data constraining long-term changes w

1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

Figure 16.12 Long-term average rainfall for southeastern Australia. Separate trend lines are shown for 1900-1995 and 1996-2006 (data source: Australian Bureau of Meteorology).



1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

in microphysical processes and the CCN environment, it is not possible to ascribe rainfall trends as being microphysically driven.

Moreover, since it is clear that climate variation over the latter part of last century in particular has led to changes in atmospheric and oceanic circulation, explanations for rainfall trends must include dynamic as well as microphysical explanations (i.e., caused by the non-stationary nature of the climate system), a proposition that we examine next.

As a lead-in to that discussion, Figure 16.13 depicts pictorially the continental rainfall trends from 1970-2006. It is evident that there are major patterns of change at work across the continent, which do not show any broad-scale qualitative relationship with population center (and hence pollutant) distribution. The decade-long national drying trend from 1996, evident in Figure 16.13, is also visible in the continental picture across the southwestern, southern, and eastern parts of the continent, as in the long-term trend to increased tropical rainfall in NW Western Australia.

Explanations of Rainfall Trends

Space limitations preclude the presentation of more than two regional examples. For the first example, we continue the discussion of rainfall trends in SE Australia, where Rosenfeld (2000) supported his claim of rainfall suppression through air pollution by alluding to a "decreasing trend in snow cover in the Snowy Mountains in the period 1897-1991" along with "decreases" (although these were acknowledged as statistically insignificant) in "snow, winter temperature and total winter rainfall." Nicholls (2000) analyzed, in detail, the average rainfall in the Snowy Mountains district (District 71), which had shown

Trend in annual total rainfall 1970-2006 (mm 10 y-1)

Figure 16.13 Rainfall trends across Australia, 1970-2006, in mm per decade (data source: Australian Bureau of Meteorology).

Trend in annual total rainfall 1970-2006 (mm 10 y-1)

Figure 16.13 Rainfall trends across Australia, 1970-2006, in mm per decade (data source: Australian Bureau of Meteorology).

a strong long-term decrease in precipitation during the period 1913-1992, and which at face value appeared to support Rosenfeld's suggestion. Nicholls showed, however, that the strong decline was artificial, that it was produced by the closure of a high altitude/high rainfall-reporting site in 1954, which meant that data from only one high rainfall site, rather than two, were contributing to the district average, leading to an artificial decline in the average over the century. His reconstruction of a composite high-elevation record showed no significant decline. This example highlights one of the major challenges in time series studies: it is imperative to ensure that time series data are of consistent quality over a decades-to-century time span.

On the separate issue of a strong decreasing trend over time, evident from the snow depths in the Snowy Mountains upon an initial observation in October, Nicholls (2005) produced another elegant analysis of the relevant data to demonstrate that the trend between 1962 and 2002 was not principally related to changes in precipitation, but rather to significant regional warming over the last half-century, which is best linked to effects of climate variation (Figure 16.14). Snow depth declined primarily because of melting (i.e., temperature), not because of a strong decrease in precipitation.

A third observation concerning rain-bearing systems in SE Australia is that only very small changes in dynamic forcing are sufficient to yield variations in precipitation. Manton (1979) demonstrated this susceptibility by detecting the orographic influence on rainfall from individual weather systems using pluviograph data from eight sites, each of which had a range in elevation of only 200 m. Manton's hypothesis was that small spatial variations in dynamic forcing

Spencers Creek snow depth

^ July-Sept. mean maximum temperature, Cabramurra Snow depth, first observation in October

Spencers Creek snow depth

^ July-Sept. mean maximum temperature, Cabramurra Snow depth, first observation in October

1962 1967 1972 1977 1982 1987 1992 1997 2002 Year

Figure 16.14 Correlation between time series trend and variability in snow depth in the Snowy Mountains, SE Australia, with time series and trend in maximum air temperature. After Nicholls (2005).

1962 1967 1972 1977 1982 1987 1992 1997 2002 Year

Figure 16.14 Correlation between time series trend and variability in snow depth in the Snowy Mountains, SE Australia, with time series and trend in maximum air temperature. After Nicholls (2005).

may induce systematic variations in rainfall from a cloud system which would otherwise be statistically homogenous. His confirmation of this hypothesis can be expected to have temporal as well as spatial implications: analysis of time series trends in precipitation over multiple decades, when evidence of an aerosol microphysical effect was sought, will need to account for the confounding causality associated instead with trends in dynamic forcing (e.g., greenhouse-induced shifts in regional dynamics, changes surface albedo, roughness, moisture availability from land-use change).

For the second example, we move from southeast to southwest Australia, where a secular decline in precipitation, on the order of 20%, has been under way since the mid-1970s. Figure 16.15 shows the associated reductions of inflow to water storages for the capital city of Perth; in February 2007, this led the State government to commission a seawater desalination plant to supply 45 gigaliters per year of potable drinking water to the city's residents.

Although it has been claimed in the Australian news media (e.g., ABC News, January 9, 2007) that the Perth rainfall decline was predominately caused by the cloud-microphysical consequences of air pollution, we are unaware of any scientific analyses or publications that support this assertion. To the contrary, extensive analyses undertaken by the multiagency Indian Ocean Climate Initiative (IOCI 2002) have identified a range of other potential contributors to the decline that are coincident with large-scale changes in atmospheric dynamics and circulation patterns. One significant argument favoring controls on rainfall in this region being large-scale is the observation that there is a very strong correlation between rainfall, at all timescales, and mean sea-level pressure (IOCI 2002). More detailed analyses show influences including changes in Indian Ocean temperature and circulation, potential effects of stratospheric ozone depletion on the Southern Annular Mode, large-scale greenhouse forcing (concluded to provide 50% of the declining precipitation signal), and land-use change (e.g., Lyons 2002; Pitman et al. 2004; Cai et al. 2003, 2005, 2007; Cai and Cowan 2006; Timball and Arblaster 2006; Timbal et al. 2006).

In addition, it is important to note that multicentury global climate model simulations suggested that the 30-year drying trend has also a finite probability of occurrence as part of natural, long-term climate variability. When analyzing any time series trend, it is therefore important to include natural variability as one component of an explanatory hypothesis. An overall analysis of the change in wintertime atmospheric states by Frederiksen and Frederiksen (2007), which used NCEP/NCAR reanalyses for the periods 1949-1968 and 1974-1994, led them to conclude that from 1975-1994 the primary process involved in rainfall reduction in SW Western Australia was a reduction of intensity of cyclogenesis and southward deflection of some storms. A poleward shift in storm tracks would appear to be consistent with the observed freshening of intermediate waters of the Southern Ocean (between 55°S and 65°S), interpreted by Wong et al. (1999) as an increase in precipitation over this oceanic region. In addition, Frederiksen and Frederiksen (2007) point out that the observed changes

Total annual inflow to Perth dams (gl)

hj in the climate of the southern Hemisphere, including reduction in equator-to-pole tropospheric temperature gradient and consequent reduction in zonal flow, are consistent with the effects of anthropogenic greenhouse forcing. However, they show appropriate restraint in stating that at this stage such a conclusion has not yet been established.

A second evaluation, this time of temporal variability in dominant synoptic types influencing SW Western Australia (Hope et al. 2006), confirmed that the frequency of occurrence of troughs associated with wet conditions had declined in the region while the frequency of synoptic categories associated with high pressure and dry conditions over the continent had increased, consistent with the findings of Frederiksen and Frederiksen (2007). A cogent summary of all these influences and a comprehensive list of relevant publications set within the context of overall detection and attribution of Australian climate change has been provided by Nicholls (2006). Notably, in this very balanced evaluation, is the acknowledgment that the Indian Ocean Climate Initiative (IOCI 2002) had considered increased local aerosol pollution as one possible causative agent for decreasing rainfall trends, but concluded that it could not be more than a secondary (unlikely) contributor in the face of evident dynamical signals. The key findings from the IOCI have been recently summarized by Bates et al. (2007), and Rotstayn (2007) provides an additional argument on the need to understand dynamical effects of pollutant aerosol.


Four key conclusions can be drawn from our Australian case study. First, it has been demonstrated beyond doubt that even for this relatively unpopulated country (on average fewer that 3 people per km2), anthropogenic emissions of CN and CCN at aggregate level dominate over natural emissions. Given the concentration of a majority of the population in fewer than ten large cities, aerosol plumes containing high concentrations of CN and CCN separated by regions of lower, or "background" CN and CCN concentration have been well known since the late 1960s. These observations are in agreement with many other studies that have been conducted in other parts of the world over the last few decades, which show increases in CN and CCN concentrations downwind of populated centers.

Second, given this situation of identifiable spatial gradients in CN and CCN levels, it has been demonstrated unequivocally that clouds forming in more polluted air contain, near the cloud base, higher concentrations of smaller droplets than clouds forming in less polluted air. Again this observation is not unique and can be generalized to other regions of the world, as has been demonstrated in many publications (e.g., Garrett and Hobbs 1995).

Third, where temporal and spatial trends in precipitation have been analyzed in Australia, changes in atmospheric dynamic processes offered the best explanations, rather than changed cloud microphysics, all other things being equal. This, too, is in agreement with other studies, such as METROMEX, which showed that although cloud microphysical properties have changed as a result of aerosol pollution, precipitation had actually increased.

Fourth, however, we must acknowledge that the work done to date is incapable of definitively ruling out microphysical forcing on precipitation efficiency as a part of the story in some circumstances (e.g., the presence of giant CCN as a way to increase precipitation development even under polluted environments). Although the observations suggest that the effect on total precipitation on the ground attributable to modification in cloud microphysics is relatively small, no experiments have been, in our opinion, yet designed and carried out anywhere that consider and link, in an integrated way, process studies to demonstrate causality, with time-series studies to define precipitation trends, while covering concurrently both microphysical and dynamic forcing on precipitation processes (i.e., all confounders at once) and their interactions. Better statistical methods need to be developed and applied to separate the relative contributions of the diffe

0 0

Post a comment