( 313 ) 



For the values of «/»,. '/', ■ • '/»,, Brons lias given tallies, so there 

 are no further difficulties about the calculations which can easily be 

 done after some practice. 



We must suffice with this short and for that reason incomplete 

 survey of the method; for further details we must refer to the above- 

 mentioned work, where all questions which may arise are discussed 

 extensively. 



3. When applying this method to observations of air-temperature 

 it has been assumed that the series need not be continued farther 

 than to the third term, so that only asymmetrical (Z)„) deviations 

 and symmetrical ones (Z) 4 ) of order one of the simple law are 

 regarded, which, with these kinds of curves not differing much from 

 the bell-shape, proves to be sufficient. When introducing terras 01 

 higher order the disadvantage moreover appears that with the evaluation 

 of the higher moments the single extreme deviations, therefore in- 

 accurately determined, play an unduly important part. As first example 

 have been selected the daily-means of the air-temperature, because with 

 these frequency-curves their obliquity changes sign along with the 

 season and can therefore be regarded as a climatological factor. The 

 daily-means are calculated from observations on temperature taken six 

 times a day during the years 1882 — 1904. 



In Table I the frequencies are given from degree to degree, 

 calculated at a total of 1000; the number of data amounts of course 

 for every month to about : 



23 X 30 = 690 or 23 X 31 = 713. 



The obliquity is immediately evident ; in winter we find extreme 

 temperatures or negative deviations which are not compensated by 

 equally large positive deviations ; in summer we find on the contrary 

 important positive deviations not contrasted by negative ones. The 

 constants of the curve, namely, the mean temperature M indicating 

 the origin of coordinates, the factor of consistency h and the coeffi- 

 cients i), and Z) 4 by which the deviations of the curve from the regular 

 bell-shape are determined are found in survey in Table II; the last 

 two quantities having reference to ;t = 1, so that they must still 

 be multiplied by 1000 for the calculation of the numbers comparable 

 to the frequencies of Table I. 



