1907.] 



Meteorological Statisth :s. 



As we are dealing with records extending over 25 years 

 each period of six weeks contains 150 values ofd ivergencies 

 from average, and each period of seven weeks contains 175 such 

 values. For each of the periods of six weeks the frequency distribu- 

 tion of the 150 values was determined and from it the 50 largest 

 positive and the 50 largest negative divergencies were selected. 

 Thus the values of the divergence from average which would 

 differentiate the " moderate " from the " heavy " and 

 " light " rainfalls were fixed. The twelve largest positive and 

 twelve largest negative divergencies were then similarly selected 

 to determine the limits for distinguishing the " very heavy " 

 and " very light " values. The periods of seven weeks were 

 treated in an analogous manner. 



The results are shown in the following table : — 





1 



II. 



III. 



IV. 



V. 



VI. 



VII. 



VIII. 



Weeks. 



2 to 7 



8 to 14 



15 t02I 



22 to 27 



281033 



34 to 40 



41 to 47 



(.8 to 1 



Divergence for — 

 Very heavy . . . 

 Heavy 

 tight 



Very light . . 



+ -So 

 + -07 



- '23 



- -41 



+ •58 



+ -io 



- '22 



-'35 



+ -6o 

 + '09 



- -23 



- -40 



+ 70 



+ '20 



- '24 



- -40 



+ 79 

 + •13 



- '22 

 -•38 



+ -So 

 + •15 



-•27 

 -•48 



+ •85 



+ "22 



- -26 



-•51 



+ 75 



+ '20 

 •27 



•4 5 



Sudden breaks would occur in the curves which define the 

 limits if these divergencies were applied to the curve of averages 

 as they stand. In order to avoid such sudden discontinuities 

 the values given in the above table were plotted on a diagram 

 in the manner shown in the lower diagram, B of Fig. 4, the 

 divergencies from average being set off as ordinates at points 

 corresponding with the central epoch of each period. The 

 points thus found were joined by a smooth curve, and the 

 divergencies for the various limits appropriate to each, week 

 were then s< t off from the curve of smoothed averages. In the 

 cases of warmth and sunshine the figures vary over a very 

 wide range and the necessity for interpolating between the 

 various periods is much greater than in the case of rainfall. 

 For instance, in District 5, "England, South," (see Fig. 3), 

 the limit for "very abundant" sunshine exceeds the average 

 by.*28 hours in the 28th week, whereas in the first week it is 

 only eight hours above it. 



