The Bainfall of South Africa. 
365 
in mind that in my tables I take the changes which have occm^ed in the 
pressure during 24 hours. 
Each morning, when the readings are plotted down on a chart, one of 
them will be actually lower or higher than the others, contingent on the 
group we have under consideration, and this condition is of importance in 
dealing with the question of prediction. We can thus divide each of our 
sub-groups into three divisions, namely : — 
(1) Where the L'Agulhas barometer has the actual lowest or highest 
reading. 
(2) Where that at the Eoyal Observatory is the lowest or highest. 
(3) Where such is the case at Clanwilliam. 
In my investigations I have taken the relative importance of the 
three stations, as far as actual pressure is concerned, in the following 
order : First, L'Agulhas ; second, Clanwilliam ; third, the Eoyal 
Observatory, and where two of these are equally lower or higher, as 
the case may be, I have given the benefit to the most important station 
of the two. 
Under Group I., with a falling barometer, the station having the actual 
lowest reading must be taken, but under Group II., with a rising baro- 
meter, the reverse will be the case. 
Again, the height of the barometer at the lowest or highest station, as 
the case may be, has a bearing on the prediction, and must be borne in 
mind. Here I have separated each division thus into sections : — 
(a) Eeadings below 29*95 in. as low. 
(b) Eeadings between 29'95 in. and 30*05 in. as mean. 
(c) Eeadings above 30*05 in. as high. 
Finally, the direction of the wind at L'Agulhas is a factor which must 
receive consideration, and one's common sense will have to be exercised 
to decide whether an existing direction of wind is likely to change or not. 
This I have been unable to go into, as it is a matter of purely local observa- 
tion and knowledge gained by experience, but I trust that some one else 
will take it up. 
In the foregoing manner I have tabulated all the triads of readings — 
which were perfect— for the years 1902, 1904, 1905, 1906, and 1907. The 
reason for omitting 1903 was that the readings from Clanwilliam were 
not taken for four months, and I did not desire to include an imperfect 
year. 
24 
