30.5 inches for the cities of Table 3. Since percentage fluctuations tend to become smaller as the total rainfall increases, the great fluctuations experienced by these tropical cities are all the more notable.

If tropical and subtropical cities having an average rainfall of less than 20 inches are included in the comparison, even more violent ranges are disclosed. For example, Cairo, Egypt and San Diego, Calif., each received about 6 times as much rainfall in their wettest year as in their driest; Athens 7 times; Helwan, Egypt, 18 times; and Onslow, W. Australia, 47 times as great.

None of the cities of Table 4, except Singapore, happens to be close to the equator, the necessary data for other equatorial cities not being readily available. However, extreme fluctuations occur almost under the equator even on oceanic islands. At Malden Island (lat. 4° 1' S.; long. 154° 58′ W.), for example, the annual totals of rainfall have varied from 3.95 inches in 1908 to 63.41 inches in 1905. At Oceanic Island (lat. 0° 52′ S.; long. 169° 35' E.), nearly 2,000 miles west of Malden Island and within a degree of the equator, the range has been between 19.61 inches in 1909 and 158.93 in 1905 (141.12 in 1911). There was likewise a range of from 74 rainy days in 1910 to 232 in 1911. Upon the Hawaiian Islands, Puuhela, on Maui, (lat. 2034° N.; long. 1562 W.) received only 2.46 inches of rain in 1912, but received 33.14 inches in 1918. Many other Hawaiian stations show a somewhat similar range, and the rainfall of the group as a whole is characterized by the government meteorologist as "extremely unreliable."

The great variability illustrated by these three mid-Pacific Islands is the more notable because insular climates are commonly

thought to be exceptionally uniform, particularly if near the equator and not dominated by near-by continental masses, nor within hurricane regions. None of these three is in a hurricane region, all are far from land and two are close to the equator.

So many other regions in low latitudes experience an undependable rainfall that it seems unnecessary to more than mention the famines produced by droughts in India and in southern China, or the destructive floods in the same countries. Tropical Australia has perhaps even worse droughts and floods and is saved from terrible famines only by the sparseness of the population and the skill used in reducing the losses to a minimum. The annual variation at Onslow in tropical West Australia, for instance, was from 0.57 inches in 1912 to 26.96 in 1900, and the average variability from year to year in that region has been about 50 per cent. of the average rainfall.

Excessive falls in short periods afford other illustrations of the uncertainty of rainfall. In tropical Australia, on more than 400 days in a 25-year period more than 10 inches of rain fell in 24 hours according to official records, while in temperate Australia there have been very few recorded instances of such heavy rainfalls-none in Victoria or South Australia and only two in Tasmania (Max. of 18.1 inches in three days). In tropical Australia, more than 20 inches has been officially recorded as falling in 24 hours on 42 different days, and more than 30 inches on four occasions. The maximum was 35.71 inches at Crohamhurst, Queensland, Feb. 2, 1893. However, 60 inches fell in three consecutive days at Mt. Molloy, Queensland, and there have been many 48-hour periods when more than 25 inches fell.

At Suva, Fiji, it frequently happens that more than 10 inches of rain falls within 24 hours; there were 4 cases in the 7-year period 1906-12. The maximum has been 26.5 inches in less than four hours on August 8, 1906.

What is believed to be the world's record for officially measured rainfall in 24 consecutive hours occurred near Manila on Feb. 14-15, 1911 (1,168 mm., 46 inches). The other stations at which this maximum has been approached are also in low latitudes, namely, Cherrapunji, India, June 14. 1876, 40.8 inches; Silver Hill, Jamaica, 57.5 inches in 48 hours; Funkiko, Formosa, 40.7 inches on Aug. 31, 1911, and at Hononu, Hawaii, 31.9 inches, Feb. 20, 1918.

With such sharp annual and daily extremes as these, it is reasonable to expect great monthly extremes. At Malden Island, mentioned above, for example, the range in officially recorded rainfall from 1890 to 1918 was as follows:

The four months, November, 1891, to February, 1892, received a total of only 0.72 inches, while the four months, January to April, 1915, received over 60 inches. The number of rainy days per year varied from 30 to 144.

At Oceanic Island, likewise, the monthly ranges are extreme. Within a nine-year period, February, March, April and November have each received 0.1 inches or less and also 21.3 inches, 28.9, 27.6 and 15.5 inches, respectively, and falls of 0.7 inches or less in May, August, September, October and December are to be contrasted with falls of from 12 to 19 inches received in other years in those same months.

The Philippines show scarcely less violent extremes. In the 16-year period, 1903-18, 42 of the 70 stations had a total of about 160 months with no rainfall, while the wettest months at about half the stations exceeded 40 inches of rain, and had less than 20 inches in the case of only 8 stations. This variation is only partly seasonal, for a month which is very dry one year may be excessively wet another. Severe and widespread droughts, with over 100 days without rain, are contrasted with destructive floods caused by rainfalls of more than 20 inches in a day or two.

Even at Hilo, on the wet side of Hawaii, where the rainfall averages 139.4 inches a year and is relatively dependable, a 13year period shows that the monthly amounts have varied widely, January from 0.5 inches to 38.6, February from 1.9 to 32.5 inches, March 2.9 to 45.4, April 3.7 to 25.1, and December from 1.7 to 27.8 inches, for example.

That the great variation from year to year in rainfall discussed in the foregoing pages is not local is suggested by various data. For example, the average rainfall of the entire Hawaiian group (150 stations) was more than twice as great in 1919 as in 1918 (112.9 in. vs. 54.5 in.). Likewise in the Philippines during the droughts such as that referred to in a preceding paragraph, nearly all of the 70 stations are affected similarly.

Another type of variation in rainfall which is prominent in the tropics is the seasonal. Very few tropical localities receive their rainfall as evenly distributed throughout the year as is common in many parts of middle latitudes. Distinct wet and dry seasons are the rule. The rainy summers and dry winters of India and China are well known. Most of tropical Australia also receives almost no rain for six months and from 15 to 50 inches or more in the other six months. Hawaii and many other places near the margins of the tropics receive much of their rainfall in winter, while still other parts of the tropics have two wet and two dry

seasons.

In order to compare the monthly variability of rainfall in low and middle latitudes, a planimeter measurement was made of Supan's map of Percentage Range of Mean Monthly Rainfall in Bartholomew's Atlas of Meteorology. This map shows four types of regions: (1) where the wettest month is less than 10 per cent. rainier than the driest month; (2) where the wettest month is from 10-20 per cent. rainier than the driest; (3) where the range is from 20-30 per cent; and (4) where it is over 30 per cent. Tables 6 and 7 show the approximate area and the percentage of each type by continents. Table 6 concerns middle latitudes (30° to 60°); Table 7 concerns low latitudes (30° N. to 30° S.).

It will be seen that low latitudes have over three times as large an area possessing a monthly variability of over 20 per cent. as is the case in mid-latitudes and twice as large a percentage of their total area has this range. The one large area in mid-latitudes having the fourth, the most extreme, type of rainfall variability is the Tibetan Plateau, which has little agricultural value because of its great altitude. Furthermore, the month of least precipitation in mid-latitudes commonly is in the winter when plants require little moisture while the wettest month usually is in summer. On the other hand, the driest month of the tropics is also a hot month, with active evaporation. This unfortunate combination is very hard on plants, and is the reason for the lack of forests in many places having a large annual rainfall. For instance, parts of tropical Australia having over 60 inches of rain a year possess no real forest because several months are extremely dry and hot.

In respect to the more uniform rainfall type, where the range between the driest and wettest month is less than ten per cent. mid-latitudes have nearly six times as large an area as low latitudes. This type comprises about 26 per cent. of the total land area of mid-laditudes while it makes up only 3 per cent. of low latitudes. Other interesting comparisons come out on further study of these tables.

Why should the lack of marked seasons in respect to temperature be emphasized and the presence of marked seasons of rainfall be largely ignored by most writers on the tropics?

Another climatic factor subject to marked changes is storminess. Cyclonic storms are erratic in all parts of the world but the extremes appear to be greatest in low latitudes. The range in the number of hurricanes damaging Australia, for example, has been from one hurricane in 1907 and 1919 to seven in 1916 and eleven in 1912. In Fiji some years have none, but several years have had three each and one year four. In the South Indian Ocean the variation reported by the Mauritius Observatory has been from one storm in 1900 to eight in 1894 (and several other years) and to thirteen in 1913. In the Philippines in a 15-year period the number of very severe typhoons varied from one in 1916 to seven each in 1908 and 1911. In respect to less violent cyclonic storms there appears to be a somewhat similar range. For example, the total number of well-marked tropical cyclones occurring in Queensland, Australia, varied from eight in 1920 to 24 in 1916. In respect to the month of occurrence, as well as in annual frequency, there likewise is marked irregularity. In some years cyclones may be lacking during the months when they normally are most frequent and occur only in months supposed to be

VOL. XV-3