Variation of Temperature in the Brit ink /s/es, &c. 73 



have minima in May and maxima late in July or early in August, and 

 the temperate winds have a very distinct minimum in November. It 

 is obvious, therefore, that the second-order effect is present in each 

 separate wind and in separate groups of winds. Nevertheless we 

 have eliminated, by separating the effect for each wind, part of the 

 physical cause which gives rise to the second-order effect. 



This is again shown in the curve which is obtained by compounding 

 the eight temperature curves (Diagram 6, fig. 2, p. 75). This gives us 

 a curve whose difference from the second-order curve* will represent 

 that part of the effect which has been eliminated by separating the 

 weather into these eight types : any effect on temperature which may 

 arise from the relative frequency with which these various types occur 

 will not have been taken into account. It will be seen that the 

 characteristics of this curve (fig. 2, Diagram 6) approach those of the 

 second-order curve far more closely than do those of the curves of 

 individual winds. It has distinct maxima in February and August 

 and minima in April and November, but the curve is not regular 

 in its period although the period is approximately semi-annual. The 

 February maximum follows a very sudden rise, and lies between two 

 values for January and March which differ but very little from the 

 minimum values in April and May ; and moreover the minimum occurs 

 somewhat earlier than the minimum of the second-order curve. 



Relative Frequency of Occurrence of the Warm, Temperate, and Cold jrind*. 



We have thus separated a part of the physical cause of the second- 

 order effect apart which depends on the direction of air supply, i.< J ., on 

 the quarter from which the air is derived. The analysis may be pushed 

 further by means of the three groups of winds which we mentioned 

 above. Curves were drawn for each group of winds whose co-ordi- 

 nates are respectively proportional to the period of the year and to 

 the percentage of the total number of days (in each month for the 

 nine years) on which the winds belonging to that group were observed 

 to prevail (Diagram 5, p. 74). It will be seen that the cold group shows 

 a very marked maximum percentage in May and a less marked one in 

 November, and minima in July and at the end of December, all of 

 which characteristics correspond approximately with the second-order 

 ciirve. The curve for the warm winds shows an inverse correspondence, 

 while the curve for the temperate winds, whose mean effect is small, 

 but whose tendency is on the whole to lower the temperature, has 

 certain points of correspondence with the curve for cold winds. It is 

 obviously impossible to represent the total share of this effect of wind 



* Fig. 1, Diagram 6, represents the total residual temperature after subtracting 

 the twenty-fifth year first-order component, and therefore shows the second com- 

 ponent combined with the residual annual component for nine years, which has an 

 amplitude of nearly '5 F. and components of shorter period. 



