SEASONAL WEATHER PREDICTION — WALKER 131 



wet years and the lower ones in dry years; in India, if the rains fail, 

 cotton and millets will grow though the ordinary crops may perish. 

 We may hope that, when our methods have improved, the predictions 

 when applied to a particular farm will be right at least three times 

 in 5 years; and if they are consistently acted upon, they will prove of 

 material value in the long run. 



Of further applications of these methods some are worthy of a 

 passing notice. For Siam, whose summer rain has a coefficient of 

 0.7 with the contemporary southern oscillation, a former Indian col- 

 league has worked out a foreshadowing formula with a relationship 

 of 0.8. And at length China, which has suffered terribly from floods 

 as well as droughts, is receiving attention. A graduate from Shang- 

 hai, now working in London, finds that the Yangtse valley and three 

 areas along the coast have enough data for a preliminary investiga- 

 tion, and has worked out formulae for prediction with coefficients 

 between 0.6 and 0.7. Mention should also be made of the researches 

 of Okada in connection with the rice crop of Japan. 



Let us now turn from the academic to the practical, and see how 

 far these theoretical methods justify themselves in actual experience. 

 I believe that the earliest regular seasonal forecasts based on meteor- 

 ological instead of astrological data were those of the Indian mon- 

 soon of June to September, started half a century ago in India by 

 H. F. Blanford, and depending mainly for their success on the ill- 

 effect upon the monsoon of excessive winter or spring snowfall in the 

 Himalayas; finally, however, he made the big generalization that 

 droughts might be associated with unusually high pres3ure over a 

 great part of Asia, at Mauritius and in Australia. Eliot continued 

 the monsoon forecasts from 1887 to 1903, but data in those days were 

 scanty; he attempted far too much detail, his mode of expresgion 

 was somewhat pontifical, and the newspapers became sarcastic; so 

 latterly he obtained immunity from criticism by printing the fore- 

 casts as confidential documents. The gradual introduction of statis- 

 tical methods in India has undoubtedly led to improvement; but, 

 as we have seen, it is much easier to predict the rainfall of December 

 to February than that of June to September, and the length of the 

 series of Indian data is not yet great enough to give complete relia- 

 bility. After careful scrutiny I estimate that of the forecasts issued 

 before the monsoon periods from 1905 to 1932 two-thirds were cor- 

 rect; but I consider that this is not good enough and that we have 

 been too ambitious. Also while the approximate prediction formula 

 of 1908 has stood the test of time with credit, the later ones of 1924 

 for northwest India and the Peninsula separately, although certainly 

 better in theory, have not, in the short period of trial, proved so 



36923—36 10 



