Meteor Streams and Rainfall Calendaricities

British Yearbook of Astronomy, 1980, pages 144-154

The two words 'meteor' (shooting star) and 'Meteorology' (the study of weather and climate) are both derived from the Greek adjective 'meteoros', meaning lofty. Because rain and shooting stars were both to be observed descending from the heavens, many people in centuries past believed that there was an association between them. However, a hundred years ago few would have expected that scientists would eventually wonder whether in fact there was not a relationship between meteors and precipitation, but in a manner very different from anything the Ancients could have imagined. To elucidate, we must firstly take a look at the process by which clouds produce rain.

The physics of rain

In order for rain to form, a mechanism is necessary by which the tiny water droplets making up clouds can be persuaded to combine together into drops large enough to fall as precipitation. Without such a mechanism, the cloud droplets would remain in the sky and eventually evaporate.

In tropical or maritime regions it is not uncommon for cloud droplets to grow into raindrops by coalescence. This will occur if a comparatively few droplets are larger than the rest, causing them to drift and therefore to sweep up and absorb the smaller ones. However, over areas of land, particularly in non-tropical latitudes, most rain commences in ice-crystal form. Because the equilibrium vapour pressure over ice is less than that over liquid water - whenever all three of these phases are present, vapour will sublime on to the frozen surfaces - while to preserve the balance, liquid water will have to evaporate. In other words the ice particles in a cloud will always grow into snow-crystals at the expense of the water drops until they are heavy enough to fall. They will then become even larger by colliding with other crystals or with water droplets, often melting into rain before reaching the ground.

It is not difficult to maintain clean water drops in a super-cooled liquid state at temperatures as low as -l5°C or even colder: the smaller the drop-size the easier it is. For this reason, many small or moderate-sized clouds with tops warmer than -15°C consist entirely of liquid droplets which remain too light to precipitate out. Ice will form only in the presence of special micron-sized aerosols, termed ice nuclei. At temperatures colder than about -25°C there are plenty of impurities in the atmosphere capable of acting as ice nuclei, but between 0°C and -l5°C there are few. In order to qualify, the surface of an ice nucleus must be compatible with the crysta!line structure of ice; however, at colder temperatures the degree of compatibility is less critical because ice forms more readily. This means that clouds penetrating to high, very cold levels in the atmosphere usually produce rain, whereas smaller clouds whose temperatures are warmer than -15°C sometimes produce negligible rainfall because of a shortage of suitable nuclei.

This reasoning has led to attempts to enhance the precipitation from moderate-sized clouds by seeding them with artificial ice nuclei (nearly always silver iodide crystals), in order to make up the deficit occurring in nature. There are grounds for arguing that the seeding not only converts many of the cloud-droplets into potential rain, but it also causes additional latent heat of freezing to be released, thus prolonging the cloud lifetime. Sometimes the quantity of extra heat energy produced may be sufficient to increase the buoyancy of the cloud to the stage where it will grow further, reaching colder regions of the atmosphere where the rain-forming process is more efficient.

Unfortunately it is far from easy to establish whether a particular seeded cloud has in fact yielded more rain than if it had been left alone. This is because of the high variability of natural precipitation: a moderately tall cumulus cell will sometimes fail to rain, whereas on other occasions it will produce a heavy downpour. Although there is agreement amongst scientists that seeding can induce changes in certain clouds, there is still debate regarding the extent to which the rainfall can be influenced.

Calendar singularities in precipitation

There is good evidence, at least in the northern hemisphere, that most natural ice nuclei in clouds consist of soil particles, such as kaolinitic clays. However, E.G. Bowen has suggested that another important source is meteoric dust from outer space. He was led to this hypothesis after noticing a number of calendaricities in rainfall - first in Australia, and then in other parts of the world. The Sydney rainfalls between 1859 and 1901 were totalled for each date in the year, and showed two exceptionally large peaks on 12 and 22 January, which departed about four standard deviations from the mean rainfall (Bowen, 1953). He then examined later records at Sydney, between 1902 and 1944, and found that prominent peaks occurred on 12/13 Janualy and 23 January, in other words on almost the same dates as in the completely different earlier period. The same behaviour was later observed by O'Mahony (1962) in four separate 25-year periods at Sydney and two such periods at Rockhampton. O'Mahony(1962) also carried out a detailed analysis of 100 years of Sydney January rainfall. He totalled up the falls recorded on each calendar date excluding first the heaviest fall on each date, then excluding the first and second heaviest, and so on. To his surprise, the residual data continued to display peaks on 12/13 and 23 January even after the ten heaviest falls of every date had been removed.

Bowen extended his investigation to other countries, examining over 100 years of records from Australia and New Zealand, 75 years from the U.S.A. and Japan, 50 years from South Africa and the Netherlands, and 30 years of data for the British Isles (Bowen 1956a and b). All seven countries exhibited rainfall maxima on or near the 12/13 January, and five of them showed a peak near 22 January. A number of other calendaricities also appeared to be too widespread to be attributed to mere coincidence, for example 18 November and 1/2 February. A subsequent paper by Brier (1961) confirmed these and other smaller Bowen singularities in American rainfall during 1952-58, a period which was not included in Bowen's study.

It seemed most unlikely that this world-wide pattern could have resulted from chance occurrences of local heavy rainfall. Bowen therefore argued that only an extra-terrestrial influence could produce nearly simultaneous effects over the entire surface of the Earth. Because the required mechanism would have to possess a yearly periodicity, he put forward an explanation in terms of meteoric particles from cometary orbits acting as ice nuclei in terrestrial clouds. Thus, Bowen associated the rainfall peak near 13 January with the Geminids meteor display of 13 December, the 23 January rain maximum with the Ursids of 22 December, the peak on 1/2 February with the Quadrantids of 3 January, and that of 18 November with the Orionids. The lag of approximately 31 days represented the time taken for the meteoric dust to settle through the Earth's atmosphere to a level where it could nucleate the potential rainclouds.

Supporting evidence for the existence of rainfall singularities came from a study of 22 years' data at six stations in the USSR by Dmitriev and Chili (1958). Of the fifteen rainfall peaks which they found between 1 August and 5 February (i.e. the period covered by Bowen), nine corresponded closely with Bowen's rainfall maxima, while four others showed up in some of the countries investigated by Bowen.

In Rhodesia, an analysis by McNaughton (1970) of the pre-1946 records of 55 stations showed eight rainfall singularities, which were confirmed by hail calendaricities in the completely different period 1957-68. All but the smallest of these singularities coincided with rainfall peaks in other parts of the world. It was possible to demonstrate that the same calendaricities were built up independently in northeast and in southwest Rhodesia from rain falling in different years (McNaughton, 1971). A similar phenomenon had been noted in a comparison of Sydney and Rockhampton rainfalls in Australia by O'Mahony (1962). This is certainly consistent with the hypothesis of a calendaricity in a secondary rain-forming factor such as atmospheric ice nucleus concentration.

By collating over 300 years of dates when snow first covered the ground at Tokyo, Bowen (1956d) managed to study a period considerably longer than most meteorological records. Calendaricities again appeared on 12 and 23 January, and on other dates when rainfall peaks had occurred (although less consistently) in his earlier investigations.

Long-period cycles in rainfall singularities

The aspect of Bowen's hypothesis which has perhaps received the most criticism is the postulated 30 or 31-day time lag between a visible meteor display and a rainfall anomaly. The meteoric dust particles cover a wide spectrum of sizes; only those with diameters of about eight microns might be expected to settle down into the troposphere in the required time, and even then it would be fair to ask whether atmospheric motion would not cause significant variations in the descent speed. Dr Bowen himself admits that there are gaps in his theory.

On the other hand it would be unfair to omit all rnention of a phenomenon which does lend at least some support to the proposed 31-day lag. This is the apparent six- to seven-year periodicity in rainfalls approximately 31 days after a meteor shower exhibiting a similar periodicity. Associated with the remnants of Biela's comet are the Bielids (sometimes also named the Andromedids after their radiant), which produce visible displays in late November and early December, The meteoric particles in the cometary orbit are concentrated in a swarm with an orbital period of about 6.5 years, similar to that possessed by the comet before it became defunct. Thus, the Bielids made brilliant appearances in 1867, 1872, 1879, 1885, 1892, and 1899, when the Earth passed through the denser section of the meteor stream, in other words at intervals of five, seven, six, seven, and seven years. (Since 1899 this shower has not-been prominent).

For the three last days of December (i.e. about a month after the Bielids' display), Bowen (1956a) examined rainfalls from the U.S.A. between 1871 and 1950, and from Australia and New Zealand between 1900 and 1950; in all three countries there was strong evidence for a six-year cycle in heavy rainfall on these dates. In the U.S.A. the correlation between 29 to 31 December rainfalls six years apart was +0.35; the probability of this having occurred purely by chance is as low as one in a thousand. There may have been a phase displacement of about a year between the rainfall peak and that of the meteor display, possibly because the particles affecting the rain were smaller than (and had gradually become displaced from) those giving the visible display. In this connection it is interesting to note that although the meteor displays have almost faded away since 1899, the effect on rainfall appears to have continued at least up until 1948.

The variations in the year when snow first covered the ground at Tokyo on 29, 30 and 31 December also showed evidence of a six or seven-year periodicity which correlated with that of the Bielids (Bowen 1956d).

The Giacobinids (or Draconids) are a northern hemisphere shower associated with the Giacobini-Zinner comet, whose orbital period is 6.6 years. The Giacobinid meteor display usually takes place on or near 9 October; it was particularly prominent in 1913, l926, 1933 and 1946. The dates of heavy November rain were extracted for 48 U.S.A. stations: it was found that the rainfalls of 8 to 10 November produced sharp peaks in 1913, 1919, 1926, 1932, 1943 and 1947 (Bowen 1956b). In other words the rain appeared to follow the same cycle as the meteor display, but about 31 days later. The rainfalls were also examined on earlier and later dates in November, but did not exhibit this periodicity.

The Perseids are another meteor stream incident on the northern hemisphere, with a period estimated at about 110 years; their activity was at a minimum in 1911. Bowen (1957) examined the records of 48 American stations on the dates 10 to 24 September (i.e. about a month after the Perseids' visible display), and found that the rainfall on these dates had decreased steadily after 1870 until (like the meteors) they reached a minimum in 1911, since when they have increased again.

Measurements of the atmospheric concentration of magnetic spherules

Whipple and Hawkins (1956) questioned Bowen's hypothesis on the grounds that the influx of sporadic meteors detected by radar and photography is of the same order as that of the regular, shower-derived meteors. However, they mentioned the possibility that this might not apply to the particles too small to be measured optically or by radar, but more likely to qualify as ice nuclei in tropospheric clouds. Bowen (1953) supports the idea that the noctilucent clouds of the ionosphere consist of meteoric dust, but the implications of this still need to be investigated more thoroughly.

During selected periods between 1967 and 1971, Rosinski (1970; R- et al., 1975) measured the concentration of tiny magnetic spherules in the atmosphere at between four and eleven different latitudes, some north and others south of the equator. On each day at every station about 1000 cubic metres of air were passed through a polystyrene filter, which was then dissolved in chloroform. The magnetic particles suspended in the solution were separated with a magnetic stirring rod, after which they were sized and counted under a microscope. Their diameters ranged between about 3 and 25 microns.

High particle concentrations appeared on similar dates at widely separated stations, proving that the particles were of extra-terrestrial origin. Sampling took place during October in three different years, all of which showed peaks on or near the seventh of that month, confirming one of Bowen's prominent rainfall singularities (Bowen, 1957). Unfortunately no other month was sampled more than once, but high concentrations of spherules were nevertheless measured on several dates near Bowen anomalies, including 9 to 14 January and 16 to 21 January. However, Rosinski (1970) mentions that Bowen's 30-day time lag is unlikely to apply to all sizes of particles collected on a particular calendaricity; in other words the differently sized spherules were probably derived from separate meteor belts.

Ice nuclei

Various methods have been devised of estimating the number of ice nuclei in the atmosphere. The most common technique makes use of an insulated cold chamber into which a sample of air is drawn. A 'cloud' will form in the chamber as the air cools by convective mixing (or by deliberately reducing the pressure); alternatively, steam may be pumped in. The number of ice crystals forming and falling out of this cloud can be estimated by illuminating them in a beam of light, or by allowing them to settle onto a glass tray. Sometimes the tray is filled with cold sugar solution which preserves and enables the ice crystals to grow, so that counting becomes easier. At -l5ºC the ice nucleus concentration of the atmosphere is often less than one per litre; at -25ºC it might be as high as 10 or even 50 per litre.

A programme of January ice nucleus measurements was organized during 1954 to 1958 at stations in Australia, America, and South Africa. The results of these and other observations have been summarized by Bowen (1956c) as well as by Kline and Brier (1958). In every one of the five years, high ice nucleus counts were recorded on or near 13, 22/23 and 31 January, agreeing with the rainfall singularities found by Bowen (1956a) and others. Furthermore, these ice nucleus peaks always appeared in at least two different continents almost simultaneously. Although there were occasions when some of the ice nucleus singularities did not show at certain stations, it is difficult to see how the apparent recurrence tendency cannot be genuine.

It is also relevant to note that many of the counts were made in aircraft at high altitude, which does not in itself prove that the nuclei were of extra-terrestrial origin, but at least it adds to the plausibility of this hypothesis. There is evidence too that the occurrence of cirriform clouds over Australia shows preferred dates on 12 and 22 January and 1 February (Bigg 1957a and b); these clouds form at very high levels in the troposphere and consist entirely of ice crystals.

In months other than January, ice nucleus counts-have been much less widespread; however, tentative support for Bowen has come from measurements made by Rosinski (1967) between December 1964 and April 1965.

Laboratory experiments by Bigg and Giutronich (1967) indicate that meteoric dust may well be capable of providing ice nuclei. Critical of other experimenters who crushed and therefore destroyed the surface properties of particles being tested, they boiled meteorites at a pressure of only 2 mm of mercury (thereby attempting to simulate conditions in the outer atmosphere), collecting the recondensed material on slides. The particles obtained from a metallic meteorite were of similar size, colour and shape to those sampled by Rosinski et al. (1975), and when tested in a cold chamber at -14ºC, they gave a count of five ice crystals per thousand spherules. A stony meteorite produced much smaller particles with a lower nucleation rate.

Discussion and conclusion

More work remains before the connection between meteor dust and rainfall can be regarded as conclusively proved. For example, there is an unexplained discrepancy between the particle concentrations measured by Rosinski et al. (1975), the nucleation rate estimated by Bigg and Giutronich (1967), and actual atmospheric ice nucleus concentrations (e.g. Bowen, 1956d). Nevertheless, as yet no other serious proposal has been put forward to account for the rainfall singularities.

Certainly there does seem to be strong evidence that these calendaricities are genuine climatic phenomena, at least during January. This was the month chosen by Bowen for the ice nucleus measurements, and it was the only month considered by O'Mahony (1962) to exhibit statistically significant rainfall peaks in Australia. January is also a month with a well-defined warm temperature singularity near 22 January in the U.S.A. (Wahl, 1953), which may somehow be associated with the rainfall singularity then (Bowen, 1956a).

Bowen suspected that his proposed meteor/rain relationship would be less clear-cut between May and September, because of greater difficulty in distinguishing one meteor shower from another during those months (Bowen, 1953 and 1957). Even in December, one of the months selected by him, the evidence for world-wide singularities is less impressive than in November, January, and early February. Bowen (1956a) mentions that the Bielids I meteor shower retrogresses in time by one day every five years, so it is pertinent to ask whether the corresponding 'favoured date' 31 days later might not also retrogress, thus questioning the value of analysing the December rainfalls accumulated on fixed calendar dates.

Another question which deserves serious consideration is the possibility of the meteoric source of ice nuclei attaining greater importance in the southern hemisphere, because of the smaller area of land and the consequent dearth of soil-derived nuclei. It is unlikely that the southern continents rely on the northern ones for a large part of their ice nucleus ration, considering the comparatively slow trans-equatorial exchange of air (McNaughton, 1971; M- and Wurzel, 1972).

Finally, even though rainfall calendaricities are almost certainly real, caution should be exercised when attempting to use them to predict weather; (the same is of course true with the recently discovered correlation between moon phase and rainfall). If the historical records of a particular area do not show evidence of a rainfall singularity known to exist elsewhere, then it may well be dangerous to infer that it will show up in that area in the future. Even when the existence of a rainfall anomaly is established using past data, such as 22/23 January at Sydney, it does not manifest itself often (O'Mahony, 1962). To illustrate, rainfall in excess of half an inch was recorded only 13 times on 23 January at Sydney between 1861 and 1960; on 58 occasions no rain fell at all. (Similar figures also apply to 22 January). This is perfectly reasonable if (as seems likely) the rainfall anomaly is associated with an ice nucleus anomaly, because the nuclei are useless unless the synoptic weather conditions are such that heavy clouds are already present in the sky.


Bigg, E.K., 1957a. 'January anomalies in cirriform cloud coverage over Australia'. J. Meteor. 14, 524-526.

-1957b. 'A new technique for counting ice-forming nuclei in aerosols'. Tellus 9, 394-400.

- and J. Giutronich, 1967. 'Ice nucleating properties of meteoritic material'. J. Atmos. Sci. 24, 46-49.

Bowen, E.G., 1953. 'The influence of meteoric dust on rainfall'. Austral. J. Phys. 6, 490·497.

- 1956a. 'The relation between rainfall and meteor showers'. J. Meteor. 13, 142-151.

- 1956b. 'A relation between meteor showers and the rainfall of November and December'. Tellus 8, 394-402.

- 1956c. 'January freezing nucleus measurements'. Austral. J. Phys. 9, 552-568.

- 1956d. 'A relation between snow cover, cirrus cloud, and freezing nuclei in the atmosphere'. Austral. J. Phys. 9, 545-551.

- 1957. 'Relation between meteor showers and the rainfall of August, September and October'. Austral. J. Phys. 10, 412-417.

Brier, G.W., 1961. 'A test of the reality of rainfall singularities'. J. Meteor. 18, 242-246.

Dmitriev, A.A. and A.V. Chili, 1958. 'On meteor streams and precipitation' (in Russian). Ak. Nauk. Mor. Gidrofiz. Inst., Trudy 12, 181-190.

Kline, D.B. and G.W. Brier, 1958. 'A note on freezing nuclei anomalies'. Mon. Weath. Rev. 86, 329-333.

McNaughton, D.L., 1970. 'Calendar singularities of rainfall in Rhodesia'. Proc. Trans. Rhodesia Sci. Assoc. 54, 99-107.

- 1971. 'Calendar singularities in Rhodesian precipitation, and the implications'. J. Appl. Meteor. 10, 498-501.

- and P. Wurzel, 1972. 'Tritium in rain as an indicator of airmass source'. Tellus 24(3), 255-259.

O'Mahoney, G., 1962. 'Singularities in daily rainfall'. Austral. J. Phys. 15, 301-326.

Rosinski, J., 1967. 'On the origin of ice nuclei'. J. Atmos. Terr. Phys. 29, 1201-1218.

- 1970. 'Extraterrestrial magnetic spherules: their association with meteor showers and rainfall frequency'. J. Atmos. Terr. Phys. 32, 805-827.

-, C.T. Nagamoto and M. Bayard, 1975. 'Extraterrestrial particles and precipitation'. J. Atmos. Terr. Phys. 37, 1231-1243.

WahI, E.W., 1953. 'Singularities and the general circulation'. J. Meteor. 10, 42-45.

Whipple, F.L and G.S. Hawkins, 1956. 'On meteors and rainfall'. J. Meteor. 13,  236-240.


Link to main index

Link to Astronomy