1099 



for the thermometric records, or Ihe incorporation of tlie latest period 

 in the historical data, might have brought about a fictitious agree- 

 ment between the curves, I pi'oceeded in the following manner. 

 Instead of all the "historical" winters only the w/-// cold or decidedly 

 mild winters (coeff. =^3 or — 3), were used for a new 89-y early 

 curve, the same weights being assigned to all and the years after 

 1827 being omitted. These data are the best, of course, but they are 

 scarce (one in 8.7 years). For further comparison another Paris — 

 de Bilt curve was plotted since 1852, now based exclusively on the 

 easily controllable deviations of the monthly means. 



These latter curves have been drawn in the middle of the diagram 

 (a and b) ; the original curves are indicated by A (Table 1) and B 

 (Table II), the more strongly smoothed curves by « and ji (Table lllj. 



TABLE 111. 



Comparison of the historical data in the 89-year period 



with the thermometric data. (Very much smoothed). 



89-y. Period 



Year of the 

 period 



Paris— de Bilt 



It seemed important to know approximately the shape of the 

 thermometric curve before 1852. I therefore marked in the diagram 

 the deviations of the monthly means (dev. from the average 1816 — 

 1860) of V2iv\%-0hservatoire for 1828—1855, by the line (little 

 crosses); although this material is decidedly inferior to that used for 

 the line B (C evidently lies too high) it still forms an interesting 

 complement. 



For the sake of brevity a, b and C have only been shown in the 

 diagram, not in the tables. 



