42 SECTIONAL ADDRESSES 
that the nature of the fluctuation so disclosed will not change suddenly, 
he forecasts that the total rainfall of some definite period, usually six 
months or a year, will be greater, or less, than it was in the previous year. 
Now a moment’s thought will make it clear that a man will in the long 
run be right three times out of four if, when last year’s rain was in defect, 
he predicts an increase, or if it was in excess he forecasts a diminution. 
So I think it is not unfair to say that success under the Swedish condi- 
tions begins at 75 percent. The success actually attained is 82 per cent., 
which is encouraging; and the success in dealing with water levels 
is phenomenally great, being slightly over go per cent. 
The seasonal conditions of Russia, which are not very closely related 
with those of the North Atlantic, have been carefully examined by W. Wiese. 
In 1923 the Hydrometeorological Office of Leningrad started publishing 
forecasts of ice in the Barents Sea, and out of seventeen monthly fore- 
casts of which I have information fifteen were approximately correct. 
Predictions of the rainfall of April and May in central and east Russia 
were initiated at the same time, and all the first four years they were 
approximately correct: the biggest difference between the actual and fore- 
casted amounts was only 20 per cent. 
No account of European activity in this department could ignore the 
enterprise of Prussia four years ago in creating at Frankfurt a.M. a post 
for research into long-period forecasting. Dr. Franz Baur has for the 
present wisely limited his activity to the issue of a forecast of ten days ; 
it would be impossible to expect results under these conditions 
which are as accurate as those of daily weather work, but I am in- 
formed that their standard fully demonstrates the trustworthiness of 
the principles employed. It is only by experiments of this kind that 
satisfactory methods of prediction can be developed. 
We may now pass to the consideration of improvements in our methods, 
and the fundamental question at once arises—what is the physical cause 
of seasonal fluctuations ? We should naturally look for it in variations in 
the energy received from the sun, and it is surprising that an increase 
in solar activity as measured by sunspots produces a slight decrease in the 
circulations in the North Atlantic and the North Pacific. In the southern 
fluctuation the tendency of numerous spots is to produce positive values, 
but even there the biggest seasonal correlation coefficient is only 0-26, 
which is much too small to provide the explanation that we seek. 
Moreover, it probably arises because a positive fluctuation is associated 
with low temperatures between latitudes 40° N. and 40° S.; and these 
are linked with an increase in sunspots. 
In order to verify that the daily pressures are not produced by short- 
lived emanations from the sun tabulations of the relationships between 
daily and weekly, as well as the monthly and seasonal, values at distant 
places have been made ; for if the daily values over the earth are controlled 
from outside there will be close parallelism between these daily and 
weekly pressures. It was found that between 31 daily contemporary 
pressures at Honolulu and Batavia the coefficient was — 0-12, which is 
negligible ; between 39 weekly ones it was +-0-10, between 47 monthly 
June pressures it was — 0-12, and between the pressures of 47 three- 
