( 315) 



of precision would scarcely afford any advantage as we are unable 

 to reduce the deviations by improving the observations. Moreover 

 the knowledge of the most probable value is of minor importance 

 as the frequency curves in general are very flat and we cannot 

 attach the common idea of errors to the deviations which, after all, 

 are more characteristic of meteorological conditions than absolute 

 values. 



Meteorological constants in the true sense of the word and to 

 which the methods and terminology of the theory of errors are 

 applicable, are nearly exclusively Fourier constants, obtained by the 

 analysis of periodical phenomena such as daily and annual variations, 

 and to these it is certainly desirable to apply the criterion of the 

 theory of errors more extensively than has hitherto been the case : 

 the theory of errors in a plane can be immediately and advantage- 

 ously used to get a clear understanding of the value of the results 

 obtained. 



If however we abandon this basis of the theory of errors and proceed 

 upon the lines whicii have of late been followed by the sociological 

 and biological sciences, the matter appears in a different light; in 

 these sciences the principal object to be obtained is not so much the 

 'mean value as the occurrence of deviations, or rather the nature of 

 the frequency curves. 



Monthly means e.g. of barometric heights may be identical for 

 January and July as far as the absolute values are concerned, but 

 we may confidently expect the frequency curves for these months 

 to bear a totally different character. It is also extremely probable 

 that the frequency curves will show a considerable difference for 

 places in different latitudes or differently situated in relation to the 

 main tracks of depressions. 



The constants which occur in the analytical expressions for these 

 curves may then be considered as characteristics of the climate and, 

 as in meteorology we possess luore data than in most other branches 

 of science, a more thorough study of details is possible. 



The principal questions are : 



a. In how far are monthly means in accordance with the common 

 law of probability. 



b. What is the form of the frequency curves constructed from 

 daily means or from observations made at fixed hours in as far as 

 these curves may be considered symmetrical. 



c. An investigation of the skewness of these curves. 



In this communication only the first of these problems will be 

 considered. 



