BAROMETRIC PRESSURES ON THE GREAT LAKES 107 



On plate 9, examine the Buffalo graphs for the period September 18-Octo- 

 ber 4. Note the extreme irregularity of the dot-and-dash graph (observed 

 elevations), showing an extreme change of 0.87 foot on October 1-2, and 

 the relative smoothness of the continuous graph (corrected elevations), with 

 no change in a single day of more than 0.12 foot. 



Plate 10 covers a stormy period, October 10-31, 1910. Note the extreme 

 irregularity of the observed elevations at Buffalo. On October 21-22 the 

 observed elevation increased 1.75 feet in a single day. The corrected ele- 

 vation shows no change greater than 0.10 feet in any single day. 



Consider the Cleveland graphs. On plates 7 and 8, though the con- 

 tinuous graph is smoother, as a whole, than the dot-and-dash graph, the 

 contrast is not great. On plate 9 the small fluctuations in the continuous 

 graph for the periods August 28 to September 10 and September 29 to Oc- 

 tober 9, 1910, are accompanied by much larger fluctuations in the dot-and- 

 dash graph. On plate 10 the same contrast, indicating much greater accu- 

 racy for the corrected elevations than for the observed elevations, shows for 

 the period October 21-31, 1910. 



As the corrections for wind effects and barometric effects at Cleveland 

 are in general much smaller than at Buffalo, it is to be expected that the im- 

 provement in accuracy produced by applying the corrections will be less 

 noticeable at Cleveland than at Buffalo, and hence that the contrast between 

 the dot-and-dash graph and the continuous graph will be much less pro- 

 nounced at Cleveland that at Buffalo. 



So, too, one may naturally expect the contrast between the two graphs to 

 be still less for the Lake Erie means as shown on the third pair of graphs on 

 plates 7 to 10 than for either Buffalo or Cleveland. Nevertheless, the 

 contrast even on these curves shows clearly in favor of the corrected ele- 

 vations. In this connection, note especially the decided smoothness of the 

 continuous graph for the Lake Erie mean for June 1-23, 1910, on plate 7 as 

 compared with the dot-and-dash graph, and the same contrast on plate 10 

 for the stormy period October 11-31, 1910. 



Compare the continuous graphs on plates 11, 12, and 13 with the dot- 

 and-dash graphs. For each of the three separate gages Mackinaw, 

 Harbor Beach, and Milwaukee it is clear that the continuous graph is 

 decidedly smoother, on the whole, than the dot-and-dash graph. In this 

 connection, attention is especially invited to the period June 23-July 15, 

 1911, shown on the lower hah of plate 11; to the period July 22-26, 1911, 

 shown on plate 12; and to the period August 28-September 19, 1911, 

 shown on plate 13. The dot-and-dash graph for Lake Michigan-Huron 

 mean is a remarkably smooth curve. It so closely approaches in smoothness 

 to the dot-and-two-dashes curve for the same lake (the rainfall+inflow 

 outflow graph) that it is necessarily difficult to tell whether the con- 

 tinuous graph is intermediate in smoothness. The graphic method of 

 comparison is not sufficiently sensitive to determine reliably the relative 

 accuracy in this case, in which both graphs, for the Lake Michigan-Huron 



8 



