CSIRO Marine and Atmospheric Research

Guidelines for the utilisation of cloud seeding as a tool for water management in Australia

Publication of the Agricultural and Resource Management Council of Australia and New Zealand

An outline of the Australian experience and principles for water managers

Brian F. Ryan
Principal Research Scientist
CSIRO Division of Atmospheric Research
Brian S. Sadler
Executive Director Water Resources
Water Authority of Western Australia
Agricultural and Resource Management Council of Australia and New Zealand

May 1995


PART I: Introduction



Legend to Figures:


In this, the driest inhabited continent on earth, cloud seeding has been carried out in various locations over some 47 years and is seen by some as a potential panacea to the devastating effects of drought. However, whilst the technology is not new, the reality is that there are only limited circumstances in which cloud seeding can be justified as an operational management investment.

Research has shown that, given the appropriate conditions cloud seeding can modify clouds and induce rain. The dilemma for water managers is that the necessary favourable conditions occur relatively infrequently and the duration of the cloud seeding experiment necessary to demonstrate increased rainfall over a given area makes any experiment costly.

Favourable circumstances have occurred, most notably in the Tasmanian hydro-power industry, where appropriate cloud conditions, effective catchment responses and positive cost benefit outcomes have justified seeding, but unfortunately, there are many more situations where these conditions do not coincide and cloud seeding is not an appropriate water management tool.

In response to the need to document the Australian experience, driven to some extent by the current severe drought in eastern Australia, Brian Ryan and Brian Sadler prepared this report which sets out guidelines for water managers, in partnership with atmospheric scientists and commercial operators, to assist them in developing planning procedures and decision-making processes that will maximise the possibility for a successful experiment.

It is recommended that wherever circumstances are favourable, a statistically designed experiment should be developed before any operational cloud seeding is undertaken because it is the only methodology available to quantitatively determine whether there is any increase from the experiment.

David Mittelheuser
Chairman, Water Resources
Agriculture and Resource Management Council of Australia and New Zealand

PART I: Introduction

The Australian continent as a whole is comparatively dry. It has many regions where limitations on the availability of water severely constrain the potential forms of economic development. However, nationally only some 15% of water resources are committed and development has concentrated in areas of more plentiful supply. Although some areas have very difficult supply situations, availability of water is not yet the critical limitation on national growth that it is in many arid or heavily populated countries.

In these circumstances a diversity of water management technologies, ranging from simple to advanced, have found their place according to the local practicalities and economics. These management approaches relate not only to supply but also to increasing end use efficiency. Within this technological and economic context, cloud seeding is one of the marginal technologies available to Australian water managers as a potential water management tool.

There are some comparatively limited circumstances, most notably in the Tasmanian hydro-power industry where the concomitance of appropriate cloud conditions, the effective hydrologic response from catchments and the positive cost benefit outcomes justify cloud seeding as an operational water management investment. There are many more situations where these conditions do not coincide and cloud seeding is an inappropriate water management tool. Even where circumstances are known to be favourable, the Australian experience shows that a statistically designed experiment is necessary before entering into an operational cloud seeding program. Currently the statistical experiment is the only methodology available to quantitatively determine any rainfall increase from seeding the clouds.

Nevertheless, despite the extent of Australian scientific experience in this field, there are still circumstances, such as in "emergency" responses to severe drought, where money is spent on cloud seeding when objective planning would indicate otherwise. This is not to say that cloud seeding would not produce extra rain on occasions, but rather to recognise that without experimental controls there can be no measure of the effectiveness of seeding, and no quantitative assessment of any extra rain or monetary benefits from the operation.

The aim of the paper is to develop guidelines that will assist water managers in determining when and where cloud seeding may be an effective management tool in Australia. In Part II of the paper there is a critical review of cloud seeding experiments that have taken place in Australia over the last 47 years. In Part III of the paper a set of desirable cloud seeding principles is developed based on the lessons drawn from Part II.


1. Background

It is often observed that supercooled clouds (ie. clouds composed of liquid water droplets at temperatures below 0°C) contain large numbers of supercooled water droplets that are too small to precipitate. The aim of rainfall enhancement techniques for supercooled clouds is to convert the supercooled water droplets into precipitation sized ice particles that then fall to ground as either rain or snow. There are two theoretical approaches to precipitation enhancement by the seeding of supercooled clouds, namely, static seeding and dynamic seeding.

The hypothesis behind static seeding is that the introduction of an "optimum" concentration of ice crystals will enhance the precipitation efficiency of a cloud by converting the reservoir of supercooled water droplets into precipitation sized particles. The static cloud seeding hypothesis has been the basis for most cloud seeding experiments in Australia.

The concept behind the dynamical cloud seeding hypothesis is that the sudden release of the latent heat of fusion when a supercooled cloud is rapidly glaciated through seeding increases the buoyancy of the cloud and this in turn generates deeper and more vigorous cloud that produces more rain. Several dynamic cloud seeding experiments have been conducted in the USA, the most notable of which was the Florida Area Cumulus Experiment, FACE (Simpson, 1980). However, extensive dynamic cloud seeding experiments have not been conducted in Australia.

A third technique for cloud seeding is that of seeding warm based cumulus clouds with hygroscopic nuclei. Rain drops in warm clouds form from the collision and coalescence of smaller droplets, that have themselves formed on cloud condensation nuclei. The hygroscopic cloud seeding hypothesis is that the large droplets that form on the seeded giant cloud condensation nuclei increase the precipitation efficiency of the updraught in the cumulus cloud. This technique has not been trialed in Australia, but is currently being seriously considered in South Africa (Mather and Terblanche, 1993).

This review of cloud seeding is focussed on the Australian experience and concentrates on the analysis of Australian static cloud seeding experiments. For a critical review of all three cloud seeding techniques in the USA the reader is referred to Cotton and Pielke (1992).

2. Historical review of past cloud seeding activities in Australia

2.1 Early CSIRO Single Cloud Experiments

Cloud seeding experiments began in Australia in 1947, shortly after the classic experiments of Schaefer (1946) in America showing that pellets of dry ice could rapidly glaciate a laboratory cloud. The first cloud seeding trials were carried out by Kraus and Squires (1947) near Sydney. In these and subsequent experiments from 1947 to 1952 scientists in the CSIRO Division of Radiophysics used Royal Australian Air Force aircraft (mostly Liberators, Beaufighters and DC3s) to drop dry ice into the tops of cumulus clouds. The conclusion reached from these experiments was that the method worked reliably and initiated rain that would not have otherwise occurred. However, the success of the rain-making was determined by the temperature of the cloud top. Below -7°C there was a 100% chance of producing precipitation, but at temperatures of -15°C and cooler the results lose their significance because of the high probability of naturally occurring rain (Bowen, 1952).

Similar trials using silver iodide smoke as the seeding agent were made from 1953-1956 in clouds located anywhere from South Australia to Queensland. Experiments were carried out using both ground based and airborne silver iodide generators. The use of ground based generators was abandoned because of concerns that the silver iodide from the generators failed to reach cloud base. Further, the nucleating properties of the silver iodide decayed during daylight hours. Efforts instead focused on the airborne dispersal of silver iodide. It was concluded that silver iodide was an effective agent in clouds whose upper levels have temperatures cooler than -5oC. Precipitation was induced within 20 to 25 minutes after seeding in cumulus clouds and somewhat later in stratiform clouds (Warner and Twomey, 1956).

2.2 Early CSIRO Area Experiments 1955-1963

During the period 1955 to 1963 four experiments were carried out in Australia in the locations shown in Fig 1 (Smith, 1974). These experimental reions were in the Snowy Mountains, in South Australia, in the New England district of New South Wales and in the Warragamba catchment area west of Sydney. The purpose in each case was to find out if rain over a specified area could be increased by seeding clouds with silver iodide released from aircraft.

With the Snowy Mountains, New England and South Australian cloud seeding experiments the time duration for a seeded period was determined by a meteorological criterion and was typically 10-15 days, while the Warragamba experiment used a 1 day period. In the Snowy Mountains experiment two areas were used as target and control respectively, and during any one period a random process determined whether clouds over the target area should or should not be seeded. In the other experiments a crossover design was chosen, that is, a random process was used to determine which area would be target and which area would be the control for each period. In other aspects the experiments were similar. Of these four experiments, only the one conducted in the Snowy Mountains produced statistically significant evidence for rainfall increases over the duration of the entire experiment with a 19% increase significant at the 5% level.

The experiments did, however, share one thing in common, namely that all appeared to record a decrease in results with time. Fig 2 shows how the initial increase in rainfall in the Snowy Mountains and New England experiments reduced in time while the rather poorer showings of the other projects became steadily worse (Smith, 1974). The situation was puzzling. Bowen (1966) postulated that the apparent decrease in effectiveness of the Australian cloud seeding operations with time was due to the persistent and cumulative effects of silver iodide on reaching the ground. More recently Bigg and Turton (1988) have argued that there is indeed evidence of prolonged increases in ice nuclei following the application of silver iodide to the ground. They have postulated that it is these secondary ice nuclei generated from the silver iodide that are involved in the persistent effects of seeding. This suggests that persistence results in rainfall enhancement that is not detected in the analysis and more importantly invalidates statistical treatments that rely on a comparison of the ratio of the target to the control on seeded and unseeded days.

In summary these area cloud seeding experiments showed that cloud seeding was not a simple technique and that it was not fully understood. CSIRO consequently embarked on a new set of area cloud seeding experiments designed to take account of the deterioration of apparent results with time, the variability of results with seeding conditions and the variability of rainfall gradients.

2.3 Seeding by State Governments

In the years from 1965 to 1971 State Governments were active in operational cloud seeding and in the ongoing debate concerning the effectiveness of cloud seeding (McBoyle, 1980). The years 1964-1966 were years of extreme drought in Australia. CSIRO set up "Courses of Instruction" in cloud seeding techniques to both inform State Government Departments and other interested parties as to what cloud seeding had to offer and to train cloud seeding officers to undertake cloud seeding operations. CSIRO acted in an advisory capacity to the states but did not carry out operational cloud seeding.

Operations or investigations of cloud seeding potential were undertaken in Victoria, New South Wales, Queensland South Australia and Western Australia (McBoyle, 1980). In all cases where the analysis of the seeding operations was possible the results were either inconclusive or, worse, controversial.

2.4 CSIRO Experiments in Tasmania

Tasmania I 1964-71.

The design of the first Tasmanian cloud seeding experiment addressed the problems raised by the previous studies (Smith, 1974). The location of the experiment is illustrated in Fig 3. The target was a Hydro-Electricity Commission catchment area located on the Central Plateau and there were three designated control areas that were unseeded on occasions when the target was seeded. This made it possible to allow for the effect of rainfall gradients in all directions. The period was divided into pairs of periods of about 12 days duration; clouds were seeded during half of the periods selected on a random basis. The seeding schedule was carried out in alternate years only. During the intervening years no seeding was performed but rainfall measurements were continued.

The experiment was carried out in alternate years until 1971 and the results suggested that rainfall increases as high as 30% were achieved in autumn at a significance level of 3%. In other words, there was only a 3% chance that the effect was not a result of the seeding. For the other seasons, the results were not nearly so conclusive, but the results for autumn and winter were sufficiently encouraging for the HEC to carry out further experiments to test the possibility of using cloud seeding as a water resources management tool. The Hydro-Electricity Commission adopted the practice, from 1972 onwards, of examining the dam levels in the major catchment areas and deciding in the light of expected water requirements whether or not to implement cloud seeding in the following autumn.

Tasmania II 1979-83.

Tasmania II was conducted by the Tasmanian Hydro-Electric Commission in the months of April to September from 1979-1983. The experiment was analysed by CSIRO (Shaw et al., 1984). The aims of the experiment were to increase the inflow into the storage lakes, to demonstrate that any increase in rainfall had not occurred by chance and to check whether areas downwind of the target area were affected. The seeding period was based on the concept of a "suitable seeding day" that was defined in advance of the experiment and a randomisation scheme based on an overall seed/no seed ratio of 2:1.

The analysed experiment showed that there were increases in rainfall following seeding and that these increases were associated with stratiform cloud in south-westerly air streams. In Tasmania II there was a 37% increase in rainfall on suitable days. The calculated increases were about half those calculated from the previous seeding experiment, but were accomplished using one tenth the seeding time.

The authors found no convincing evidence of effects of seeding down wind of the target area. However, they noted that this may have been because the analyses were not sufficiently sensitive to detect any such effects. More recent analysis (Searle, pers. comm.) also suggest that there is no evidence of the effect of seeding down wind of the target area.

Tasmania III 1992-1994

The Tasmania III experiment had a design similar to that of Tasmania II the main difference being that the seeding agent was dry ice and not silver iodide. The three year experiment (1992-1994) conducted and analysed by the Tasmanian Hydro-electric Commission was run from May to October. The meteorological and flight data are currently being prepared for detailed statistical analysis. In addition to the Tasmania III cloud seeding experiment, drought relief projects were undertaken over the Central Plateau catchments in 1988, 1989, 1990 and 1991 and over the east coast and midlands of Tasmania in 1994. These drought operations were not randomised so any evaluation of these activities was confined to historical analyses. However Searle (pers. comm.) finds that the historical analyses give results that are consistent with those obtained in the earlier experiments.

2.5 CSIRO Emerald Experiment 1972-1975

Studies of large cumulus clouds were made in the area around Emerald, Queensland from 1972-1975 (Fig 1). The clouds were considered to be suitable for seeding in a region where the increased rainfall would have been valuable for irrigation and mining. In a series of experiments, isolated cumulus clouds were seeded using both silver iodide and dry ice. When dry ice was used as the seeding agent, the seeding produced an almost immediate conversion of the cloud tops from water to ice and precipitation echoes were observed on the 3cm aircraft radar. The same effect was not observed when silver iodide was the seeding agent. However, the problem of detecting extra rain on the ground was complicated by the extreme variability of the rainfall in the area studied. Intense, widely separated showers fell from the isolated cumulus clouds or cloud masses, a situation that made for difficulties in the measurement of mean area rainfall and necessitated the use of large numbers of raingauges. As a result of these preliminary experiments it was concluded that while opportunities for cloud seeding undoubtedly existed in the area, a controlled experiment would need a duration of many years if it were to yield a reliable answer.

2.6 CSIRO Experiment in Western Victoria 1979-1980

The Western Victorian cloud seeding experiment was carried out in a major wheat-growing area where extra rain during the growing season would be of economic benefit (Fig 1). The planning for the experiment involved a new degree of sophistication in cloud seeding experiments in Australia. There were three pre-conditions associated with the site selection.

These were:

Clouds suitable for seeding had to occur in an area reasonably frequently, and be identified with a particular synoptic-scale weather pattern, namely, rain depressions or "closed lows".

The frequency of occurrence of these suitable clouds, and the amount of extra rain expected to be derived, had to be such that there was a reasonable chance of detecting the seeding effects in five years.

The cost of mounting the experiment had to be considerably less than the economic benefits from the extra rain.

Cloud observations, statistical analysis of past historical weather records and economic analysis of wheat yield as a function of rainfall were undertaken from 1975-1978 and in 1978 a prospectus giving full details of the experiment was circulated internationally to interested scientists (King et al., 1979). In this way the experiment received international scrutiny, and the integrity of the analyses was preserved by publishing the analysis procedures beforehand.

The analyses showed that approximately five "closed lows" were expected each season. Although five per season was considered to be a low number of seeding opportunities, the rainfall from these systems was uniform, producing good correlations between target and control area rainfalls.

Despite this very careful planning, during the 1979 and 1980 seasons there were only seven days that satisfied synoptic conditions defined by the experiment and of the synoptically suitable days when optimal seeding conditions were expected no suitable conditions were found. This was due almost entirely to the abundance of natural ice crystals in the cloud system (King, 1982).

In retrospect it is apparent from the 1979 and 1980 field data that the number of seeding opportunities was overestimated in the pre-experiment studies based on the analysis of synoptic charts. A re-analysis of historical data showed that 60% of the closed lows with deep stratiform clouds had cloud top temperatures colder than -25oC. In-cloud observation suggested that these clouds invariably contained more than enough ice crystals for precipitation to form. The re-analysis showed that realistic expectations for the seeding of closed lows were fewer than 1.5 per season compared to the initial estimate of five per season thereby giving no prospect of obtaining answers on the effects of seeding in the originally planned five-year experiment (King, 1982).

At the same time as the seeding project was being re-evaluated, a revised economic analysis showed that aircraft costs were increasing relative to wheat prices The analyses suggested that by 1983 the cost/benefit ratio for the experiment would have been entirely eroded (King, 1982). It was therefore clear that continuing the Western Victorian cloud seeding experiment could not be justified.

2.7 Western Australian Northern Wheatbelt Cloud Study 1980-1982

The Northern Wheatbelt Cloud Study was undertaken in Western Australia by the Western Australian Weather Research Association (WAWRA) and the Western Australian Government (Bailey, 1982). The operational seeding and research program was managed primarily by the School of Physics and Geosciences at the Western Australian Institute of Technology (now Curtin University). The study ran from 1980 to 1982 during the months from May to October. The location of the project is shown in Fig. 1.

The cloud physics study showed that in general cloud tops were relatively warm (-5 to -15°C). Large numbers of ice crystals were more frequent in May and June, the first two months of the observations when the rainfall was greatest. When liquid water was present in the clouds it frequently exceeded 1g m-3. However, on the best seedable days the rainfall was usually light and at best represented 29% of the total rainfall in the season.

The three year study showed that while there were a number of clouds each year that had a reasonable potential for seeding, simulation studies showed that a 30% rainfall increase was needed to have a reasonable chance of detecting rainfall in a five year experiment. To detect a 10% increase in rainfall would require a 20 year experiment.

An economic analysis of a seeding experiment suggested that the cost of an experiment would be $3.31 per hectare of land under cultivation with an initial $0.89 per hectare in the first year to cover capital costs. If a 10% increase in rainfall was achieved on all the best seedable days that occurred in 1981 and 1982 the extra yield per hectare was worth between $5 and $17 per hectare.

It was concluded that in the northern wheatbelt region of Western Australia is probably suitable for a cloud seeding operation. However, the duration required for a cloud seeding experiment made the area unsuitable.

2.8 Melbourne Water Corporation/CSIRO Experiment (1988-1992)

The Melbourne Water Corporation conducted a five year cloud seeding experiment (1988-1992 inclusive) from 1 May to 31 October over the 487 km2 Thomson Reservoir catchment in the Great Dividing Range about 120 km east of Melbourne (Fig. 1). Melbourne Water contracted CSIRO to oversee the design and conduct of the experiment and to make measurements of the physical properties of clouds in the area.

Two kinds of clouds were seeded in the experiment. Stratiform clouds with and without embedded cumulus were seeded the requisite distance upstream for the induced precipitation to fall into the target area. Cumulus clouds were seeded a fixed 30 minutes travel time upwind of the target area.

Suitable clouds for seeding were selected according to criteria based on cloud top temperature, cloud depth and appearance, cloud supercooled liquid water, wind velocity and the likelihood that clouds would live long enough to be seeded during two passes along the seeding track.

Stratiform or cumulus clouds upwind of the target area were seeded with silver iodide in acetone solution, while dry ice pellets were dropped into orographic cap clouds and cumulus clouds with tops warmer than -5°C. The choice to seed or not to seed clouds with silver iodide was made randomly with a 2:1 balance in favour of seeded experimental units. For dry ice the balance was 1:1.

Two major airborne field studies were conducted in 1988 and 1990 as an adjunct to the cloud seeding experiment. The field studies showed that the post frontal clouds behind frontal systems and the inter-frontal clouds between frontal systems offered the best seeding opportunities (Long, 1993).

Observations during the Melbourne Water cloud seeding experiment confirmed the hypothesis that substantial quantities of liquid water could at times be generated by orographic uplift when the cloud systems were relatively shallow (cloud top temperatures warmer than -10°C). The results of analyses of the rain gauge network showed that any increase in rainfall was not statistically significant. However, other tests for the buffer area between the target and the control areas showed a statistically significant increase.

3. Critical assessment of past cloud seeding activities in Australia

In Australia, hydro-electric authorities and water boards have looked to cloud seeding to supplement reservoirs for hydro-electric power and for water supply for public and industrial consumption. Currently many utility groups are either negative or doubtful about the effectiveness of cloud seeding. In a survey in 1979, McBoyle (1980) reported a negative response from the State Rivers and Water Supply Commission of Victoria, Electricity Commission of New South Wales, and the Queensland Water Resources Commission. Earlier Sadler and McCullogh (1969) reported that Western Australia's initial attempts at rainmaking were not successful (McBoyle, 1980). The more recent cloud seeding experiment by Melbourne Water was also deemed to be statistically inconclusive and consequently seeding operations have not been pursued.

The only mainland study to show a statistically significant increase in rainfall was the Snowy Mountains experiment (1955-1959). However, a dispute arose between the CSIRO and Snowy Mountains Hydro-Electricity Authority (SMHEA) over the weight to be placed on the results of subsidiary analyses. To the SMHEA they suggested that the increase due to seeding was substantially less than the 19% analysed by CSIRO while to CSIRO the limitations of the subsidiary analyses appeared such as to prevent the drawing of meaningful conclusions (Smith et al., 1963). At the insistence of SMHEA, the final report on the experiment (Smith, Adderly and Walsh, 1963) interpreted the significance levels of the results to be "marginal" and the overall results of the experiment as "encouraging" but "inconclusive". The reason the experiment is viewed as unconvincing is because, (i) changes were made after the first year, (ii) the discovery that the precipitation values for the experimental units were calculated by people who were aware of the randomised sequence and (iii) the dispute between CSIRO and SMHEA over subsidiary analyses.

The lesson to be drawn from the Snowy Mountains experiment is that extreme care needs to be taken in the statistical design and conduct of cloud seeding experiments. In Australia, these lessons were well learned. The basic statistical rules required to design and evaluate a cloud seeding experiment developed for the Tasmanian experiments have been accepted internationally by statisticians.

It is perhaps ironical that in 1993 SMHEA circulated an environmental impact statement to assess a proposal to undertake a 6 year precipitation enhancement project, as part of an overall technical feasibility study of cloud seeding over the Snowy Mountains region (Harasymiw and McGee, 1993). Both the CSIRO Division of Atmospheric Research and the Bureau of Meteorology were invited to make submissions on the proposal. CSIRO advised against commencing the experiment until the methodology to evaluate the seeding was peer reviewed and more completely documented.

The Tasmanian Hydro-Electric Commission is convinced of the economic success of the Tasmanian experiments. This is perhaps best illustrated by the decision of the HEC to undertake the Tasmania II experiment without any operational assistance from CSIRO. However, the HEC has retained a very pragmatic approach to cloud seeding. McBoyle (1980) quoting from Watson (1976) states that "Cloud seeding has emerged as a feasible and economic proposition in Tasmania when the increase in precipitation can be utilised for power generation". However, it is still viewed as a marginal benefit and its inclusion in the power generating system presents a number of managerial, design and operational problems. Currently Searle (1994) estimates that each HEC cloud seeding operation costs $645,000 to run and returned an average 55 mm of extra rain each 6 months experimental season. When the extra water in storage is priced against the energy generated by the only HEC thermal station the real profit from the silver iodide seeding averages out at about $14.5 million per annum (Searle, pers. comm.)

In times of severe drought, affected communities have called for operational drought-relief seeding. CSIRO’s role in cloud seeding has been determined by the Federal Government to be one of applied scientific research. Within this framework CSIRO cannot become directly involved in operational cloud seeding simply because no scientific research is associated with such exercises. The role of CSIRO has rather been a consultative one in providing the expertise required to perform the seeding most efficiently while at the same time warning that no quantitative results can be determined by statistical or other known methods.

Some state authorities have undertaken short term seeding programs. For example, in November 1994 the NSW State Government funded cloud seeding in northern New South Wales for drought relief. Cloud seeding was carried by the Tasmanian Hydro-Electricity Commission. Other State Government Departments sought advice from CSIRO on the viability of cloud seeding as a drought relief option.

It is perhaps ironical that in recent times the major criticism of the design of cloud seeding experiments comes not from the statisticians but from a cloud physicist, Dr E.K. Bigg. He has questioned the fundamental design of past cloud seeding experiments in Australia. Based on the arguments initially proposed in Bigg and Turton (1988), Bigg (pers. comm.) considers that all of the Australian area cloud seeding experiments were contaminated by the persistence effect of silver iodide on the ground thereby underestimating the effect of seeding. If the claims are true, new statistical techniques will need to be developed for the design of cloud seeding experiments.

4. Circumstances where rain enhancement experiments are not favourable in Australia

King (1982) examined the relevance of the western Victorian cloud seeding experiment to other areas of Australia. The following two sections showing that rainfall enhancement is unlikely to be effective for winter and spring over the inland plains of southern and eastern Australia and for summer rainfall over plains of eastern and north eastern Australia are taken from King (1982). In addition to the regions deemed unfavourable for a cloud seeding experiment by King (1982), the study by Bailey (1982) suggests that the region of Western Australia immediately to the north of Perth is also unsuitable for a cloud seeding experiment.

4.1 Winter and spring rainfall over the inland plains in southern and eastern Australia

The western Victorian site was chosen for a cloud seeding experiment because its meteorology and agriculture was representative of a large fraction of south-eastern Australia The conclusion from the western Victorian experiment is that the following synoptic systems are generally unsuitable for seeding, (i) frontal systems (including pre- and post-frontal), (ii) south-westerly airstreams, (iii) closed lows and deep cloud systems (King 1982). Table 1 shows the percentage of the daily rainfall associated with each synoptic system for the months of July to November for 10 years from 1960-1970 together with the total rainfall for Beulah, a rainfall station in the target area for the western Victorian experiment. The systems not examined during the western Victorian experiment were troughs, north-easterly airstreams and unclassified systems. These represent 10.8% of the climatology and could be assumed to be potentially seedable until proven otherwise. However these represent only 9.5% of the season’s rainfall.

King (1982) extrapolated the results of the western Victorian experiment to other inland plain regions in Australia by undertaking a similar analysis of the synoptic rainfall patterns for stations located at Maitland in Yorke Peninsula (SA), Wagga Wagga (NSW) and Gunnedah (NSW). The locations of these sites are shown in Fig 1. Table 1 shows for each location the daily rainfall associated with one of six synoptic weather patterns and the fraction of the total season’s rainfall. The table shows that for Maitland there is virtually no opportunity for rainfall enhancement and this is consistent with the earlier seeding experiments by CSIRO from 1953-1963 in this region. Near Wagga Wagga on the western plain of NSW only 11% of the season’s rainfall comes from potentially suitable synoptic systems. At Gunnedah the seedable opportunities increase to 37.8% of occasions and these account for some 36% of the rain. As one proceeds further north, the summer tropical weather patterns (troughs and north-easterly streams) dominate more and more and the higher latitude systems (fronts and closed lows) become less important. Gunnedah has an even balance between summer and winter rainfall and is at the northern limit of the area for which the meteorological conclusions derived from the western Victorian experiment can be extrapolated.

The analysis of King (1982) shows that the inland plains of South Australia, Victoria and New South Wales are not particularly suitable for seeding. On the economic side King (1982) examined the price movements of beef, wool and cereals. With all of these commodities the rises in aircraft operating costs outstripped the returns to farmers by a factor of two from 1972-1980 and indeed it was impossible to find any measure of agricultural return appropriate to the inland plains that had not declined in relation to aircraft costs.

4.2 Winter and spring rainfall over the northern wheat-belt region of Western Australia

In the northern wheatbelt of Western Australia there is a prima facie case for seeding clouds with rainfall enhancement potential and furthermore, from an agricultural view point seeding of these clouds would be economic. From a Water Management view point the potential targets are not in catchment regions. The disappointing aspect of the study is that any increase in rainfall is unlikely to be detected in a 5 year cloud seeding experiment. These results are entirely consistent with the lack of conclusive evidence from an earlier seeding operation over the catchments to the south of the Darling Scarp (Sadler and McCulloch, 1969).

4.3 The summer rainfall regions of northern Australia

Although the incursion of a tropical cyclone into the area can dominate the whole rainfall season, these are comparatively rare and on average very little of the season’s rainfall comes from large-scale synoptic systems with widespread stratiform rain. Instead, a large fraction of the rainfall events in these regions involves cumulonimbus clouds giving rise to extreme spatial and temporal variability in the rainfall. In terms of detecting seeding effects the spatial variability necessitates the deployment of many hundreds of automatic raingauges or the use of a sophisticated radar which can fill gaps in the rainguage network. Simulations using the technique developed by Twomey and Robertson (1973) show that in northern Queensland the increased rainfall variability requires four times as many opportunities compared to western Victoria to detect a seeding increase of the same magnitude. This means that an experiment that would produce a statistically significant result in south-eastern Australia in 5 years would take 20 years in northern Queensland. Consequently, the planning of a cloud seeding experiment in the summer rainfall regions of northern Australia requires a commitment of massive resources to an extremely long experiment and this is very difficult to justify on economic grounds.

5. Circumstances where rain enhancement experiments might be utilised in Australia

The region where cloud seeding seems to be effective is in orographic regions where the flow over the mountains substantially enhances the rainfall. However, Australian cloud seeding experiments in regions of rapid orographic uplift have met with mixed success.

In Tasmania three cloud seeding experiments have been undertaken. Both experiments indicate increases in rainfall in autumn and winter. The Tasmania II experiment shows that the majority of seeding opportunities occurred with cloud tops between -10°C and -12°C. At temperatures colder than -15°C there were very few opportunities for seeding (Shaw et al., 1984). The most suitable clouds are associated with stratiform cloud in a maritime south-westerly airstream. Seeding of these cloud systems resulted in a 37% increase in rainfall. Suitable days occurred 18 times a year during the experiment and this gave rise to an estimated total increase of 197 mm for seeded days.

Cost/benefit analyses carried out by the Tasmanian Hydro-Electric Commission for the Tasmanian I experiment suggest that the increased rainfall from seeding represents a gain of 13:1. More recently Searle (1994) argues that the three separate cloud seeding projects sponsored by the Hydro-Electricity Commission of Tasmania spanning 14 years have confirmed that cloud seeding can routinely enhance runoff into Tasmanian storages by 10-20%. Searle estimates that the energy gained by the cloud seeding operation cost less than 0.2 cents per kilowatt hour.

In contrast to the Tasmanian cloud seeding experiments, the Melbourne water experiment showed no statistical increase in rainfall over the catchment area. The reasons for the rainfall increase in the buffer area on seeded days is not understood and it is not clear if this can be attributed to the effects of seeding.

King shows that for the Snowy Mountains only 100 mm or 11% of the season’s rainfall occurs in the shallow south-westerly stream with cloud top temperatures warmer than -15oC. Even if increases similar to Tasmania occur, this represents only an increase of 37 mm or a 4% increase in total rainfall. Analyses undertaken by the Snowy Mountains Hydro-Electric Authority (Harasymiw and McGee, 1993 or Warburton and Wetzel, 1991) show that clouds containing the maximum liquid water contents occurred near -9oC and were associated with shallow orographic cloud consistent with King’s earlier analysis. The Snowy Mountains Hydro-Electric Authority assume a more substantial increase of 10% for their proposed new experiment. Given this interpretation of the cloud physics it is likely that the inconclusive results found by Melbourne Water would be repeated. However, this pessimistic interpretation is weakened if the earlier CSIRO Snowy Mountains experiment is accepted as being statistically significant. A statistically significant increase would almost certainly imply that the simple static cloud seeding hypothesis was not that driving the rainfall enhancement process.

The effectiveness of cloud seeding in the New England District of NSW, like that in the Snowy Mountains, is still controversial. A re-analysis of the earlier New England experiment using daily periods rather than the initial 14 day seeding periods suggests a substantial increase in rainfall over all years (Bigg, pers. comm.). However, this re-analysis is not consistent with the published design for the experiment, and therefore will remain controversial. More recently, Chambers and Long (1992) completed a precipitation enhancement feasibility study in the region of the Copeton Dam, which is sited in the New England District of NSW. They concluded that there is a good potential for cloud seeding in the Copeton Dam region, with the odds of 1:4of detecting a 30% increase in rainfall in a 5-6 year experiment and about 1:2 for a 20% increase.

A major difference between the meteorology in the New England region and the regions further to the south is that the clouds with seeding potential in the New England region are cumulus and not stratiform clouds. Indeed there is evidence based on aircraft observations made during a recent drought relief project to suggest that the deep rain bearing stratiform cloud systems are unsuitable for cloud seeding (Searle, pers. com.).

6. New developments in cloud seeding technique

Since CSIRO ceased active research into cloud seeding there have been several significant developments that have not yet been applied in Australia. These technologies are reflected in two fields that have undergone rapid development in the last 10 years, namely, new instrumentation development and the application of numerical modelling techniques to cloud seeding.

Radars with Doppler and polarising capability are being used to track both the in-cloud properties and air motions in seeded and unseeded clouds. Satellite imagery is being used to distinguish between clouds containing ice and those containing water. Microwave radiometers are being used remotely to measure the liquid water contents of clouds. The synoptic and mesoscale networks needed to analyse the cloud systems being seeded are being enhanced by wind profilers, automatic mesoscale surface networks and automatic raingauge networks. The telemetering of these data to a central control office and to the aircraft enhances the seeding operation.

While these new measuring systems undoubtedly enhance the physical interpretation of the cloud seeding experiment, they are expensive to install and require expert technical support to maintain. In Australia currently there is only one research aircraft and one liquid water radiometer suitable for cloud seeding studies. There are no airborne or mobile Doppler radars available for cloud studies. Consequently, in new Australian cloud seeding studies any options for applying these new technologies will need comparatively large budgets to fund both the operation of the instruments and to support the scientific expertise required to operate them and analyse the results.

Numerical modelling techniques have advanced to the stage where high resolution numerical cloud models are able to simulate the generation of rain from cloud systems. They are extremely useful for testing cloud seeding hypotheses, for calculating the trajectories of ice crystals to devise aiming strategies, to ensure that any seeded rain falls into the target area, and finally to calculate whether or not the seeding material will enter the cloud as hypothesised.

The greatest drawback in the use of numerical models for cloud seeding studies is that the theoretical basis for ice generation in clouds is still incomplete. Until this knowledge gap is completely filled, model simulations will not be able to be used to verify the increased rainfall from a seeding experiment.

7. Hygroscopic methodologies of cloud seeding

Recently, research in South Africa has suggested that hygroscopic flares can be effective in modifying the warm rain process in clouds (Mather and Terblanche, 1993). Cloud seeding using a hygroscopic seeding agent is not new and has been applied in several countries including India and Thailand. However, as pointed out by Cotton (1984) the major problem in establishing the efficacy of warm cloud precipitation enhancement has been that randomised field experiments do not clearly establish that an observed precipitation anomaly is physically linked to seeding.

Recently, a new approach to hygroscopic seeding has been developed based on observations made in both the USA (Eagan et al., 1984) and South Africa (Mather, 1991) that precipitation anomalies occur down wind of the plume of a Kraft paper mill. Airborne observations show that large anthropogenic nuclei are transported into the cloud base of cumulus clouds by updrafts where they act to promote the early formation of drizzle drops while inhibiting smaller cloud condensation nuclei from becoming cloud drops. The net effect is a broadening of the initial cloud droplet spectrum and an early initiation of the coalescence process.

A South African flare was designed to create this effect artificially in cloud and has been used in trials in South Africa in a randomised hygroscopic cloud seeding experiment (Mather and Terblanche, 1993). The randomised experiment is based on an analysis of the radar returns from seeded and unseeded clouds. The analyses were done for 49 clouds observed in the summer of 1991/1992 (29 clouds were seeded with hygroscopic flares while 20 were left unseeded). The statistical study showed that the seeded clouds generally grew larger than the unseeded clouds and that the seeded cloud had a larger rain mass for clouds with lifetimes greater than 20 minutes. By the end of the 1993/1994 summer season, the statistical population had increased to 97 and confidence in the result had improved markedly. The mean rain mass of seeded storms was significantly larger than the mean mass of unseeded storms at the 3% and 2% significance level respectively for the 40-50 minute and 50-60 minute intervals after the seeding decision. The South Africans consider that the experiment reached acceptable levels of statistical significance in a single season.

However, the experiments have yet to undergo the ultimate test of an objectively designed area experiment that deliberately attempts to produce more rainfall on the ground.

8 Conclusions

Over the last 47 years successive cloud seeding experiments and microphysical investigations of the clouds have shown that the static ice crystal cloud seeding hypothesis is not effective in enhancing winter rainfall over the plains area in Australia. However, there is evidence to suggest that cloud seeding is effective for limited meteorological conditions in stratiform clouds undergoing rapid orographic uplift. In Tasmania, there is strong statistical evidence for rainfall enhancement for clouds with tops between -10°C and -12°C in a south-westerly airstream. The evidence for similar effects on the mainland in the vicinity of Melbourne, the Snowy Mountains and the New England District of New South Wales is less convincing. However, the increased rainfall may be underestimated if the hypothesis of Bigg and Turton (1984) that the target area analyses are affected by persistence is correct.

In the summer rainfall regions of northern Australia, the extreme rainfall variability makes it impossible to design a statistical experiment that is able to be evaluated in a reasonable time period using the currently available techniques. Rainfall enhancement experiments in these regions remain inconclusive. Over the inland plains of Western Australia, the seeding opportunities are too infrequent to permit a realistically funded cloud seeding experiment. This is not to say that cloud seeding would not produce extra rain in these regions, but rather to recognise that currently there is no acceptable technique to demonstrate the effectiveness of seeding, that is, the extra rain and monetary benefits from the operation are not measurable.

Apart from Tasmania the prospect of cloud seeding based on the simple static hypothesis of cloud seeding seems to be very limited. This is particularly so when the requirements of a statistically significant result in 5-7 years with demonstrable economic returns are demanded in the experimental design.

In the Australian context cloud seeding based on either the warm cloud seeding hypothesis or the dynamic cloud seeding hypothesis has not been properly evaluated. The recent South African cumulus cloud experiments using hygroscopic seeding are promising but are based on single cloud studies. However, these rainfall increases have yet to be established in an area rainfall experiment. Given the Australian single cloud seeding experience, the extrapolation of the single cloud results to area cloud seeding experiments must be treated with extreme caution.

Economic analyses based on the Tasmanian experiments, Melbourne Water experiments and the proposed Snowy Mountains experiment suggest that for water management purposes increases of the order of 5-10% in the rainfall makes a cloud seeding operation economically viable. However the analyses by King (1982) suggest this is not the case for the agricultural sector that includes wool wheat, beef and sugar. In Western Australia, the analysis by Bailey (1982) suggests that cloud seeding is economic for cereal crops.

Given the current priorities for atmospheric research in CSIRO, the study of weather modification techniques must compete with funds for research into climate change, climate variability and air pollution studies. It is likely that any substantial research into this area in the future will be initiated by the Water Industry and will require substantial support from that body. However CSIRO will retain its expertise in the fundamental cloud physics necessary to evaluate any studies undertaken by the Water Industry.

9. Acknowledgments

This section of the paper could not have been written without the support of former members of the CSIRO Division of Cloud Physics. In writing the section concerning the statistical design of cloud seeding experiments, I have been guided by Doug Shaw from the CSIRO Division of Mathematics and Statistics, while Part II of the paper has drawn heavily on an internal CSIRO report written by Warren King. I have appreciated comments on the manuscript from Jack Warner and Keith Bigg.


1. An overview

In this section of the paper guidelines for the utilisation of cloud seeding as a tool for water management in Australia are recommended. The guidelines are based on the experience gained from the 47 year history of cloud seeding in Australia and have been developed to aid planning and decision-making for water managers in effective partnership with atmospheric scientists and commercial operators.

The guidelines are developed as principles that recommend the disciplines to be followed in the planning and implementation of a cloud seeding experiment that seeks to maximise the opportunities for defining and achieving a successful outcome. The central theme is one of a planned approach, with clear accountabilities and quality assurance, proper understanding and realistic performance objectives and measurements.

Throughout this document four main parties to an operation or experiment are identified and their roles discussed. These parties are:

Water Manager

Design Scientist

Cloud Seeding Operator, and

Independent Review Scientist.

In particular circumstances the roles may vary somewhat, but important separations must remain. Thus, for example, there should be a clear role separation between Commercial Operator and Design or Review Scientist. However, there may be circumstances where it is appropriate for the Design Scientist and Operator may come from the same oganisation.

Throughout this document it is assumed that the rainfall enhancement operations should proceed under guidance of a documented plan, the production of which will be managed by the Water Manager as client.

The Design Scientist is contracted by the Water Manager for his meteorological and statistical expertise to give objective advice on the seeding potential of the proposed project, to design the project incorporating best international practice so as to ensure the cloud seeding project will adequately address the objectives and provide a measurable outcome that can be evaluated by the Water Manager.

The Cloud Seeding Operator is responsible for all operational aspects specified in the project plan and is contracted by the Water Manager.

An Independent Review Scientist is contracted by the Water Manager in the planning stages of the project to ensure that the proposed project meets best international practice and that during its conduct the integrity of the project is maintained. At the end of the project an Independent Review Scientist is again contracted to assess the results of the experiment. The two contracts need not necessarly be undertaken by the same scientist.

2. Principles

2.1 Regional Evaluation

As a background to any specific program of rainfall enhancement, it is desirable to identify meteorological regions that are likely to be favourable to cloud seeding, and those regions where cloud seeding is likely to be unfavourable. The cloud seeding experiments discussed in Part I of the paper serve as a starting point for a climatology of clouds suitable for seeding with silver iodide.

The majority of Australian cloud seeding experiments have been designed on the so called static cloud seeding hypothesis, namely that excess supercooled liquid water in the clouds can be converted into precipitation by seeding the clouds with a substance that generates ice crystals. In nearly all cases complexes of silver iodide have been used as the primary seeding agent.

In Australia, area cloud seeding experiments suggest the following regional circumstances are unfavourable for effective cloud seeding by the static cloud seeding method.

Winter and spring rainfall over the inland plains in southern and eastern Australia

Winter and spring rainfall over the wheat-belt region of south-western Australia

The summer rainfall regions of northern Australia

The regions are unsuitable either because the clouds are unsuitable or because suitable cloud conditions occur relatively infrequently. A region also may be unsuitable even when suitable clouds occur but the variability of the rainfall requires an excessively long cloud seeding experiment to demonstrate any effects of seeding by the statistical techniques currently available.

Where air flows over mountains substantially enhancing precipitation, the Australian experience has been that cloud seeding is successful at some locations, while at other locations the effect was either not statistically significant or, worse, the results were controversial. For example:

Tasmania (HEC) - successful

Baw Baw Plateau (Melbourne Water) - no statistically significant increase

Snowy Mountains (CSIRO) - controversial

The last case is an example of an experiment that was successful on the statistical design criterion. The controversy arose because of a dispute over the subsidiary analyses and the unfortunate circumstance of the rainfall analysts being aware of the randomisation sequence. In the case of the latter criticism, the results were unchanged when re-analysis was commissioned from a group unaware of the randomisation sequence.

A major unresolved source of controversy concerning the Australian experiments concerns the existence or otherwise of persistence effects arising from the use of silver iodide. However, no substantial scientific resources have been committed in CSIRO or elsewhere to find an alternative seeding material for static cloud seeding or to study if indeed persistence effects enhance precipitation.

From the water management point of view, persistence may be a beneficial phenomenon. However, until a physical mechanism for persistence is accepted by the meteorological community at large, the internationally acceptance of a statistical methodology for the design and validation of a cloud seeding program that assumed persistence would be controversial and the interpretation of the success or otherwise of the experiment would be disputed.

In the Australian context, dynamic seeding and hygroscopic seeding techniques have not been used in regional cloud seeding experiments and therefore the effectiveness of these techniques in Australia is unassessed. The concept behind the dynamical cloud seeding technique is that sudden release of latent heat of fusion when a supercooled cloud is rapidly glaciated through seeding increases the buoyancy of the cloud and this in turn generates a deeper and more vigorous cloud that produces more rain. The hygroscopic cloud seeding technique introduces giant condensation nuclei into the bases of cumulus clouds. The hypothesis is that the large droplets that form on these seeded giant cloud condensation nuclei increase the precipitation efficiency of the updraft in the cumulus cloud.

While overseas experiments have shown both dynamic and hygroscopic seeding techniques to effectively modify individual cloud systems, they have yet to be shown to increase rainfall in a statistically designed area experiment.

The undertaking of a dynamical or hygroscopic cloud seeding experiment requires both significant scientific expertise and substantial resources. Any cloud seeding experiment designed to use either of these techniques would draw heavily on overseas studies and would need to be accompanied by a major research program to develop the appropriate tools and methodologies to evaluate these techniques under Australian conditions. Since the techniques are aimed at enhancing the precipitation efficiencies of cumulus clouds, they are most likely to be successful in a summertime rainfall regime.

2.2 Regional assessment of potential for rainfall enhancement to produce beneficial water management outcomes.

In Australia the areas with the most favourable potential for rainfall enhancement to produce beneficial outcomes are where suitable climatic conditions are juxtaposed with high value water developments that are already at full utilisation.

Those areas currently recognised as showing more favourable potential for specific evaluation are:

hydro-electric projects in Tasmania and the south-eastern highlands

growing urban water supplies in similar climatic circumstances

Water Managers should treat the development of any precipitation enhancement project as a four stage process:

The pre-planning of experiments and operations

Design of the experiment

Conduct of the experiment

Evaluation of results

These principles for designing a rainfall enhancement experiment have been employed in all recent projects undertaken or supervised by CSIRO. They include the recommendations of the World Meteorological Organisation to Government Decision Makers (WMO, 1986).

2.2.1 Stage 1: Pre-planning

The elements of the pre-planning are set out below and would be undertaken by the Design Scientist for the Water Manager. These steps may be pursued iteratively with subsequent phases of refinement dependent on justification or otherwise. They are most likely to overlap the design phase of the experiment. In the negotiation prior to the contract between the Water Manager and the Design Scientist it is important to:

Identify Key Participants, their roles and partnerships

Agree to the essential contents required within the project plan

Assess meteorological and cloud conditions and develop an hypothesis for rainfall enhancement

Define rainfall enhancement objectives and goals

Define expected water management outcomes

Establish a benefit/cost assessment for expected range of possibilities

The last three points are closely linked so that the process of defining the objectives and goals of the project, the water management outcomes and the economic assessment would not be developed in isolation.

Identify Key Participants, their roles and partnerships

At this stage the Key Participants are the Water Manager and the Design Scientist, together they should develop the contents of the project plan that identifies the skill, resources and budget required to undertake a cloud seeding experiment.

Agree to the essential contents required within the project plan

In developing the project plan the Design Scientist needs to have the technical expertise to assess the meteorology and cloud conditions and to provide the Water Manager with a hypothesis for rainfall enhancement. The Design Scientist would define rainfall enhancement objectives and goals based on the seeding hypothesis. The Water Manager would convert the potential rainfall enhancement outcomes into water management outcomes and establish a benefit/cost assessment for the possible range of outcomes and then would decide whether or not to proceed with the design stage of the project plan.

Assess meteorological and cloud conditions and develop an hypothesis for rainfall enhancement

The assessment of the meteorological cloud conditions requires, at the very least, a study of the climatology of the region based on archived Bureau of Meteorology surface and upper air analyses, radar studies and satellite imagery. In addition special observing and numerical modelling studies may be required to validate the seeding hypothesis.

Suitable clouds should be defined based on the seeding hypothesis. A preliminary analysis should define the frequency of occurrence of the suitable clouds, and the wind speed and direction at the seeding level for such clouds. These parameters contribute to the definition of the target and control areas for any experiment.

Define rainfall enhancement objectives and goals

Statistical techniques should be applied to determine the probability of detecting a range of seeding outcomes in a given time period. The seeding outcomes should be based on the physical seeding hypothesis. Statistical simulation and the seeding hypothesis should be the basis for defining the necessary length of the cloud seeding experiment to detect the expected rainfall increase.

Define expected water management outcomes

The Water Manager should be responsible for defining the desired water management outcomes from the project and with the aid of the Design Scientist assessing practical benefits to be pursued. The Water Manager should also assess any potential or expected negative environmental or economic effects.

Establish a benefit/cost assessment for expected range of possibilities

The Water Manager should undertake a cost/benefit assessment to determine the economic viability of the project and seek a review of this assessment by the Review Scientist. The Water Manager should assess the worth of the project on the basis of the pre-planning analyses in consultation with interested parties who may be favourably or unfavourably affected. If the project is to proceed the Water Manager should carefully define its experimental and water management objectives in conjunction with other interested parties.

2.2.2 Stage 2: Design of Experiment

Once the decision to proceed with the cloud seeding experiment is made by the Water Manager, the Water Manager and Design Scientist should proceed to Stage 2 (the Design of the experiment) using the following steps:

Document project plan

Design of statistical evaluation methods and consequent project evaluation

Establish specifications for cloud seeding operations and monitoring

Identify environmental and social issues of concern

Arrange independent scientific review of the plan

Prepare and award contract

Document project plan

The project plan or prospectus should restate the defined objectives, expected outcomes (benefits and negative impacts). It should determine the design of the experimental side of the project and define the operational areas and suitable days for the experiment. The experimental design should be based on a statistical analysis that uses past records from an existing raingauge network. These rainfall records should be continuous over a period of the order of 30 years. The experiment should define only a small number of key hypotheses to be tested so as to avoid the problems that arise from a multiplicity of statistical significance tests. The seeding technique and pluviograph/rain gauge networks should be specified together with other specialised measurements, such as extra radiosondes. Seeding and research aircraft flight patterns should be presented and finally the methodology to be used in the statistical analyses and physical assessment documented.

The cloud seeding and cloud observing requirements in the project plan determine the minimum specifications for a cloud seeding aircraft and if ground based seeding is to be used, the positioning of the seeding generators should be specified. As a precursor to locating the generators, it is important to carry out numerical simulations of the dispersion of the plume from the generator to show that the seeding material will enter the clouds at the required location for effectively seeding the target area.

The role and location of remote sensing equipment such as a liquid water radiometer or operational radar would be defined together with the techniques to be used in the analysis phase of the experiment.

The operational application of numerical modelling techniques would be specified and in particular the role of atmospheric numerical models in deciding the suitability of a seeded day, the targeting the rainfall and in the analysis of the experiment should be clearly documented.

Design of statistical evaluation methods and consequent project evaluation

Comparison of precipitation during seeded periods with that during historical periods presents problems because of climatic and other changes from one period to another. Problems also arise because of changes in the instrumental recording of the data. For these reasons historical analyses are not generally accepted as a technique to validate a seeding experiment.

Historically, statistical analyses of cloud seeding experiments require the establishment of target and control areas in the design of the experiment together with a rigorously defined analysis technique that is specified prior to the commencement of the experiment. Currently, randomised experiments are considered the most reliable technique for detecting cloud seeding effects. The randomised experiments require a number of cases that can be calculated on the basis of the natural variability of the precipitation, the magnitude of the expected effect and the certainty required in claiming the positive effects of seeding. The reliability of the randomised experiments is crucially dependent on both the quality and the length of the historical rainfall records.

The methodology for assessment of hydrological outcomes and the translation of these outcomes into an assessment of a cost/benefit evaluation should be determined and specified in the plan. The efficacy of the cost/benefit analysis will in part be determined by the availability of historically reliable economic and hydrologic data sets.

Establish specifications for cloud seeding operations and monitoring

The prospectus should define the preliminary procedures to be undertaken by the Cloud Seeding Operator both before takeoff and on the way to the seeding area. The Cloud Seeding Operator should examine all the relevant information about conditions including forecasts, reports from previous seeding or reconnaissance flight, relevant satellite and radar imagery. On the way to the seeding area the flight plan should allow the Operator to observe both cloud conditions and winds at various heights on climb.

The seeding equipment and seeding material should be specified in the prospectus together with the aiming strategies for the clouds being seeded. For example the technique of seeding stratiform clouds differs from cumulus clouds. The prospectus should also deal with what is done on occasions declared "unseeded".

Identify environmental and social issues of concern

The prospectus should address environmental and social issues that may concern the public. For example, the prospectus should comment on the impact of the seeding operation on the physical and biophysical environment as well as the likely down wind effects on communities outside the seeding area. The prospectus should state the environmental safeguards and monitoring employed. In some cases a Referee may be appointed with the power to stop the experiment if excessive rain is likely to be detrimental to the seeded area.

Arrange independent scientific review of the plan

The prospectus should be reviewed by an independent Review Scientist to help ensure that all the procedures are in accordance with developing world practice and science and to ensure the integrity and credibility of the analyses is preserved by publishing analysis procedures beforehand.

2.2.3 Stage 3-Conduct of rainfall enhancement operations

Following the decision of the Water Manager to proceed with the cloud seeding operation, a contract should be let with the Cloud Seeding Operator. During the conduct of the experiment it is important that the Cloud Seeding Operator follow the procedures of:

Project management and reporting

Seeding procedures

Cloud physics measurements during the operation

as defined in the science plan for the cloud seeding project. These procedures should not be changed without the agreement of the Water Manager, Design Scientist and the Project Review Scientist.

The science of statistics has developed rules for the conduct of experiments. Violation of these rules by the Operator will reduce the credibility of the results of the analysis of the cloud seeding operation. The rules generally are directed to the task of obtaining an unbiased set of data. Consciously or unconsciously, persons working on the project may have a bias regarding its results. This will affect the quality of the data and the subsequent analysis and therefore precautions in the design of the experiment are taken to exclude this bias. Consequently, the Cloud Seeding Operator must maintain high standards of reporting to the Water Manager so as to ensure that the integrity of the experimental procedure is never questioned.

The successful treatment of any suitable cloud requires that sufficient quantities of the appropriate seeding material enter the cloud in a timely, well-targeted fashion. The stringent spatial and temporal targeting specified in the seeding procedures must be followed by the Operator at all times. To undertake these tasks professionally the Operator needs to have a commitment to understand the science of the project.

Cloud physics and meteorological measurement must be made during the seeding operation to ensure that seeding takes place under the optimum physical conditions as defined by the seeding hypothesis. The statistical analyses are enhanced if physical predictors are employed in the statistical analysis of the seeding experiment.

2.2.4 Stage 4-Post operational evaluation

The post operational evaluation should:

Re-assess the meteorology of the cloud systems in the light of observation made during the cloud seeding operation

Evaluate the statistical success or otherwise of the experiment

Carry out explanatory analyses of data from the experiment

Evaluate the water management and economic outcomes

Disseminate the results of the evaluation of the project

The overall post-operational analysis together with recommendations should be prepared in a final report by the Design Scientist for the Water Manager. Before accepting the final report the Water Manager should submit the final report to the Review Scientist for an independent assessment.

Re-assess the meteorology of the cloud system in the light of observations made during the cloud seeding operation

The meteorology of the cloud systems should be re-analysed using all of the available data collected during the life time of the project. The data should be analysed in such a way as to document the types and frequency of clouds suitable for seeding.

Where possible, the physical evidence both for and against the seeding hypothesis should be examined. The physical analysis should take account of all of the available observational and modelling information. The analysis may be complex and require measurements made by a seeding aircraft, a dedicated cloud physics research aircraft and remote sensing equipment such as satellite imagery, radars and liquid water radiometers.

The information from the meteorological and cloud physics analysis should be used to initialise cloud models to further test the seeding hypothesis.

Evaluate the statistical success or otherwise of the experiment

The small number of analyses specified in the prospectus as being the determinants of the experimental outcome should be carried out. These will be directed towards confirming the seeding hypothesis by detecting, with an agreed level of statistical significance, increases in the area rainfall measures on specified occasions. The number of such analyses should be small (ideally one or two) to avoid statistical problems arising from multiplicity of statistical significance tests.

The analyses performed should possess high statistical power to ensure maximum ability to detect the expected effects and should be robust against unexpected features of the data.

Carry out explanatory analyses of data from the experiment

The Design Scientist should carry out extensive analyses of data from the experiment, extracting maximum value from the large amount of data on rainfall, meteorology and cloud physics that will have been accumulated. These analyses are virtually unlimited in scope, but it should be emphasised that they do not contribute to any statistical assessment of the experiment, which flows only from the pre-specified analyses.

Explanatory analyses may be directed at:

examining and elaborating aspects of the seeding hypothesis

looking at subsets of the data, either in space or time

carrying out analyses suggested by the features of the data actually obtained

improving the selection of suitable conditions and seeding strategies for obtaining the maximum effect of seeding in the designated target area in future operations

Evaluate the water management and economic outcomes

The hydrological outcomes of the project should be assessed in relation to the project objectives and in accordance with procedures stipulated in the project plan.

The economic outcomes of the project should be assessed in relation to the project objectives and in accordance with the procedures stipulated in the project plan and the water management outcome evaluations should be developed in consultation with the parties interested in the potential benefits and impacts of the experiment.

Disseminate the results of the evaluation of the project.

The post operational analysis of the statistical, water management and economic outcomes of the project, together with the basis of the assessment should be presented in the form of a final project report.

This report should set out valid conclusions to be derived from the project with respect to:

whether the project objectives were fulfilled

any new knowledge acquired from the project relating to the operational worth of further rainfall enhancement programs or to the regional science of rain enhancement

any recommended further actions or suspension of actions in respect to rainfall enhancement

After review by the Review Scientist the report should be published in a suitable form and distributed with the objectives of:

satisfying the needs of parties with a direct interest in the project outcome

providing accountability for the operation

adding to regional knowledge on the potential benefits, difficulties or failure of rainfall enhancement as a water management tool

3. Conclusions

Many years of research in Australia have shown that, given the appropriate conditions cloud seeding can modify clouds and induce rain. The problem for Water Managers in applying the technique as a water management tool is that the conditions for cloud seeding to work occur relatively infrequently and the duration of the cloud seeding experiment necessary to demonstrated increased rainfall over a given area may be excessively long thereby making the experiment too costly. History has also demonstrated that deficiencies in the statistical design of a cloud seeding experiment generate controversy and lead to inconclusive results.

In the Paper we have suggested that the only region in Australia where the static cloud seeding hypothesis is likely to be effective is where the air is undergoing orographic uplift. There is strong evidence to suggest the methodology is effective in Tasmania. However, on the mainland the claims of success have been either controversial or inconclusive.

Area cloud seeding experiments based on alternative cloud seeding hypotheses, such as the dynamic cloud seeding hypothesis or the warm rain hypothesis have not been used in Australia and have yet to have been shown to be effective anywhere in world. However, the encouraging results from South Africa were noted.

It is important that Water Managers not view cloud seeding as a proven technology that works in all circumstances, but rather view it as a technique that is effective in only a limited number of especially suitable meteorological conditions. Where a Water Manager perceives cloud seeding may be effective in augmenting water supplies the principles outlined above should assist in the planning and in the implementation of any cloud seeding experiment. The central theme of the principles is one of a planned approach, with clear accountabilities and quality assurance with proper understanding, realistic performance objectives and measurement.


Bailey, I.H., 1982: Northern Wheatbelt Cloud Study Report. Final Report Western Australian Institute of Technology, 991pp.

Bigg E.K. and E.Turton, 1988: Persistent effects of cloud seeding with silver iodide. J. Appl. Meteorol., 27, 453-457.

Bowen, E.G., 1952: Australian experiments on artificial stimulation of rainfall. Weather, 7, 204-209.

Bowen, E.G. ,1966: The effect of persistence in cloud seeding experiments. J. Appl. Meteorol., 5, 156-159.

Chambers, L.E. and A.B. Long, 1992: Precipitation enhancement feasibility study in aid of cotton irrigation. Report to DPIE Cotton Research and Development Corporation, Narrabri, NSW, Project No CSA1C, 31pp.

Cotton, W.R., 1984: Modification of precipitation from warm clouds-A review. Bull. Amer. Met. Soc., 63,146-160.

Cotton, W.R. and R.A. Pielke 1992: Human impacts on weather and climate. ASTeR Press, 288pp.

Eagan, R.C., P.V. Hobbs and L.F. Radke, 1984: Particle emissions from a Kraft paper mill and their effects on the microstructure of warm clouds. J. Appl. Meteorol. 13, 535-552.

Harasymiw, B and J. McGee, 1993: Snowy Precipitation Enhancement Project: A proposal to evaluate the feasibility of increasing snow precipitation over the Snowy Mountains area. Snowy Mountains Hydro-electric Authority, Cooma, NSW, Australia.

King, W.D., M.J. Manton, D.E. Shaw, E.J. Smith and J. Warner, 1979: Prospectus for a cloud-seeding experiment in western Victoria. CSIRO Division of Cloud Physics, CP 222, 48pp.

King, W.D. 1982: Cloud-seeding in Australia: Experiments 1948-1981. CSIRO Division of Cloud Physics, CP 326, 17pp.

Kraus, E.B. and P.Squires, 1947: Experiments on the stimulation of clouds to produce rain, Nature, 159, 489-491.

Long, A.B., 1993:Supercooled liquid water and precipitation enhancement opportunities in Australian winter mountain storm clouds. Aust. Met. Mag., 42, 17-30.

Mather, G.K., 1991: Coalescence enhancement in multicell storms caused by emissions from a Kraft paper mill. J Appl. Meteorol., 30, 1134-1146.

Mather, G.K. and D.E. Terblanche, 1993: The national precipitation research programme. Water Research Commisssion Report, Pretoria, South Africa. 133/8.93, 123pp.

McBoyle, G.M., 1980: Weather modification: Australia’s role in the world scene. Dept. of Geography, Univ. of Queensland, 183pp.

Sadler, B.S. and R McCullogh, 1969: Cloud observation and seeding operation, Interim Report. Public Works Department, Perth, Western Australia.

Searle I.L.,1994: A method of increasing catchment runoff. 7th Australian Cotton Conference, Queensland, Australia, Cotton Growers Research Association, 591pp.

Shaw, D.E., W.D. King and E. Turton, 1984: Analysis of Hydro-electric Commission cloud seeding in Tasmania, 1979-83. CSIRO Division of Cloud Physics, CP 394, 33pp.

Schaefer, V.J., 1946: The production of ice crystals in a cloud of supercooled water droplets. Science, 104, 457-459.

Simpson, J., 1980: Downdrafts as linkages in dynamic cumulus seeding effects. J. Appl. Meteor., 19, 477-487.

Smith, E.J. E.E. Adderly and D.T.Walsh, 1963: Cloud-seeding experiment in the Snowy Mountains, Australia, J. Appl. Meteorol., 2,324-332.

Smith, E.J., 1974: Cloud seeding in Australia. Weather and climate modification, Ed W.D. Hess, Wiley&Sons, NY, 842pp.

Twomey S. and I. Robertson, 1973: Numerical simulation of cloud seeding experiments in selected Australian areas. J. Appl. Meteorol., 12, 473-478.

Watson, B., 1976: A review of cloud seeding in Australia and its potential impact on water resources management. Proceedings of the 5th workshop of the United States/Australia rangelands panel, Bolse, Idaho. 181-192.

Warner, J. and S. Twomey, 1956: The use of silver iodide for seeding individual clouds. Tellus, 8, 453-457.

Warburton, J.A. and M.A. Wetzel, 1991: Field study of the potential for winter precipitation enhancement in the Australian Snowy Mountains. Atmos. Res., 28, 327-363.

WMO, 1986: Information concerning weather modification directed to Government Decision-makers. WMO/TD 123, Geneva, Switzerland.

Legend to Figures: (link to images)

Figure 1: Location of cloud seeding experiments in Australia. E, (Emerald Experiment), NE, (New England Experiment), W, Waraggamba Dam Experiment, SM, (Snowy Mountains Experiment) MW, (Melbourne Water Experiment), WV (Western Victorian Cloud Seeding Experiment), SA, South Australian Experiment, WA (Western Australian Northern Wheatbelt Cloud Study) and Tasmania (T). Also shown are the towns of Gunnedah (NSW), Wagga Wagga (NSW), Thredbo (NSW), Beulah (Vic) and Maitland (SA).


Figure 2: Variation of observed result with time in each experiment. The ordinate is rainfall increase factor associated with seeding. (from Smith, 1974)


Figure 3: Experimental areas in Tasmania (from Smith, 1974)

Site updated 1st October 2003

Modified: April 3, 2008

Link to Copyright Email CSIRO Enquiries Link to Legal notices and disclaimer Link to Privacy Statement Link to CSIROnline