384 



MR. S. M. JACOB ON 



This, of course, will produce a flatter curve system than would otherwise be the 

 case. 



The curves obtained from the monthly data are superposed. 



4th. Taking the average for the 20 years 1887-8 — -1996-7, and taking 20 day 

 periods we have the following data : — 



Rain in -^ths inches. 

 20th May— 8th June . . . . . . 6 



9th June — 28th June 

 29th June — 18th July 

 19th July — 7th August 



8th August — 27th August 

 28th August — 16th September 

 17th September — 6th October 



7th October — 26th October 



24 

 64 

 88 



99 

 47 

 13 



1 



These give the average ' monthly ' rates, that is, the amount of rain which would 

 fall in one month, if the given rate for the 20 days were maintained throughout the 

 month. 



This scheme has been adopted because the monthly rainfall is the most fre- 

 quently tabulated datum, and comparison becomes easy. 



In order to fit the above with a skew curve of type y =y (i-\--\e a % we & n & 



the moments about the mean to be 



/* 2 = 1-675 

 ** 3 = - * 2 74 



We have £ 2 = r6, P % - 3 =14 with a probable error of -18, so that the skewness is 

 significant, and Gaussian curve will not sufficiently represent the data. 

 The equation of the curve is 



x \ 249-4 



y = 105-66 (1 \ 



J \ 20-40/ 



I2-23*. 



The distance between mean and mode = 392. 



This gives a modal value for the 5th August. 



The points of inflexion of this type of frequency curve are given by 



x= + 



-V p 



■-, so that in this case the points of inflexion are 26 days on either side 



of the mode, that is, at the nth July and the 1st September respectively. 



If we remember that the south-west monsoon begins and ends with a certain 

 abruptness, these dates will roughly determine the typical duration of the monsoon, 

 though, of course, the limits may often be exceeded. 



Figure 22 shows the curve in question. 



