1098 



TABLE II. 



Winter-temp. Paris— de Bilt combined, 1852—1916. 



year for year. It may be asyumed that tlie so obtained t h e r m o- 

 metric curve is in the main fairly representative of the 

 course of winter temperature from 1852 to 1916 in the area dealt 

 with. 



The data for the historical and the thermometric curve, both 

 smoothed by C— | {a -{- 2h -\- 'Sc -\- 2(1 + 2e), are found in tables 

 I and II. 



I first verified whether some neighbouring periodicity (e.g. of 88, 

 90, 88'/, years) could improve the similarity l)et\veen the two 

 curves ; a simple graphical method sufficed for this. The result was 

 negative ; only a period of 88 '/, years gave in part of the curve 

 a somewhat better i-esult ; there appeared nevertheless to be sufficient 

 reason for preferring the 89-year period. 



In order to meet the objection that the averaging process used 



