The Bainfall of South Africa. 
379 
DIVISION (3). 
Sectio7i (a). — Out of 7 cases, only 2 were followed by rain ; wind was 
easterly. 
Section (b). — Here we have 11 cases, rain following only after 1 of 
them. 
Section (c). — Out of 10 cases, only 2 were followed by rain, and then 
not more than 0*02 in. fell. 
Prediction for Division. — Dry weather. 
By applying the result of the foregoing investigation to the rainfall 
returns of 1908, I arrive at the following : — 
Under this group there were 191 cases, 15 of which I included as 
failures, being 7*8 per cent. But inasmuch as upon 5 occasions our 
common sense would have made us relegate some of the cases to other 
sub-groups, we get the total number of absolute errors, or failures, as 10, 
making 5*23 per cent, of the total number of cases. I think this is proof 
enough of the value of the argument. 
GEOUP II.— GENEEAL PEESSUEE INCEEASING. 
SUB-GEOUP (A). 
By examining the triangle, it will be seen that when pressure increases 
at L'Agulhas more than at the other two stations, it is either due to an 
anticyclone passing to the south of the Cape or to the fact that increase of 
pressure generally is retarded by summer conditions to the north. This is 
more noticeable when Clanwilliam has the least increase. Sometimes the 
barometer at Clanwilliam actually falls when the other two are rising, 
and this brings the condition into relationship with that under Sub- 
Group 1, (G). 
In this Sub- Group the actual highest pressure bears a direct in- 
fluence on the following weather ; the normal condition is a rising 
barometer at each station, with the actual lowest reading at Clanwilliam 
and the highest at L'Agulhas ; such denote pure summer type conditions, 
and it depends entirely upon the height of the barometer whether these 
conditions are due to a depression to the north or an anticyclone to the 
south. Should the actual reading be lowest at L'Agulhas, it is evident 
that winter type conditions are making themselves felt, possibly a depres- 
sion has just passed with a very low reading at that station, so that the 
excess of increase in pressure is only apparent ; in such a case rain would 
be probable. In all predictions this condition must be taken into account, 
and if it exists the probability of rainfall should be mentioned. 
