T. Russell — Prediction of Cold-waves. 473 



with a weight of one. The mean temperature of a section is 

 the mean of its bounding isotherms. When the decrease of 

 temperature towards the northwest in a distance of 500 miles 

 is not more than thirty degrees, the greatest fall may be taken 

 as not greater than five-sixths of the change in 500 miles. 



The probable error of greatest temperature-fall by this rule 

 as derived from 201 cases of cold-waves is about ±2/5 degrees 

 for falls of twenty degrees and ±6*5 degrees for falls of fifty 

 degrees. Table III shows the distribution of errors. 



Table III. 

 Errors in Computed Temperature Falls. 



ises. 



Error in degrees. 



Cases. 



Error in degrees, 



14 



0° 



5 



±10° 



36 



±1 



4 



±H 



24 



±2 



3 



±12 



22 



±3 



7 



±13 



14 



±4 



3 



±14 



19 



±5 



1 



±15 



19 



±6 



4 



±10 



10 



±7 



1 



±17 



15 



±8 



2 



±18 



8 



±9 







'-J-The computed fall is less than the observed. On the aver- 

 age it is less by one degree for falls of twenty degrees ; by two 

 degrees for falls of thirty ; by half a degree for falls of forty ; 

 and by five degrees for falls of fifty degrees. The average of 

 all cases gives the computed fall two degrees less than the 

 observed. 



This method of deriving the maximum fall can only be used 

 where the pressure gradient is at least 04 of an inch in 500 

 miles. It is worthy of note that the mean temperature 

 throughout the areas of high and low pressure, as derived 

 from the planimeter measurements of areas between isotherms, 

 agrees in most cases within a few degrees with the tempera- 

 ture at the place of greatest fall after the fall has occurred 

 as computed by this rule. Where the computed falls differ 

 greatly from the observed, probably correspond to times of 

 widely differing diminution of temperature upward in the air. 

 With the computed maximum fall derived in this way, and 

 the computed extent of temperature- fall, the areas of ten and 

 twenty-degree temperature-falls can be taken from suitably 

 prepared tables. The agreement of the computed and observed 

 areas is on the whole tolerably satisfactory. It is difficult to 

 estimate the accuracy of the method without a map for each 

 special case. There are two cases where the method fails 



« Am. Jour. Sci.— Third Series, Vol. XL, No. 240.— Dec, 1890. 

 30 



