106 EFFECTS OF WINDS AND OF 



Hence, apparent fluctuations in the graphs of observed elevations or of 

 corrected elevations which exceed the rates stated in the last sentence of the 

 preceding paragraph are evidence that said graphs are imperfect represen- 

 tations of the rise and fall of the mean surface of the lake. The nearer a 

 given graph approaches to the smoothness of the dot-and-two-dashes graph, 

 or to the rates of fluctuations stated in the last sentence of the preceding 

 paragraph, the more accurately does it represent the fluctuation of the mean 

 surface of the lake. If one examines the graphs of observed elevations and 

 of corrected elevations on plates 7 to 13 with this point of view, it becomes 

 clearly evident that the corrected elevations are much more accurate than 

 the observed elevations. 



Examijne first the Buffalo graphs on plates 7 to 10. 



Even for June 1 to July 15, 1910, a comparatively quiet part of the year, 

 when the wind effects and barometric effects at Buffalo are much smaller 

 than in October, note how much smoother is the continuous (corrected 

 elevation) graph than the dot-and-one-dash (observed elevation) graph on 

 plate 7. Note, for example, the following contrast: On June 1-4 the ob- 

 served elevation decreased 0.67 foot in three days, immediately increased 

 0.50 foot in the three days June4-7,then decreased 0.51 foot in the three days 

 June 7-10, and then increased 0.52 foot in the two days June 10-12. The 

 changes in the corrected elevations in these same successive periods were 

 respectively a decrease of 0.12 foot, an increase of 0.11 foot, no change, and a 

 decrease of 0.15 foot. The fluctuations in the corrected graph were not more 

 than one-fourth as rapid as on the observed graph in the interval June 1-12. 

 During the period June 1-July 15 the greatest change in the observed ele- 

 vation (dot-and-dash graph) in any single day was 0.50 foot decrease on 

 July 3-4. During the same period as covered by plate 7 the greatest change 

 in a single day in the corrected elevation, as shown by the continuous graph, 

 was a decrease of 0.20 foot June 25-26. It is evident that the corrected 

 elevations at Buffalo do not represent the fluctuations in the mean surface of 

 Lake Erie without error, for there are days on which the corrected elevation 

 changes more rapidly than it is possible for the elevation of the mean surface 

 of Lake Erie to change. But it is equally evident from the graphs that the 

 corrected elevations are a much more accurate representation of the fluctu- 

 ations of the mean surface than are the observed elevations. From an 

 inspection of plate 7, one may conclude that the errors in the corrected ele- 

 vations are from one-half to one-quarter as large as those in the observed 

 elevations. 



Consider next the Buffalo graphs on plate 8. In general, one finds the 

 same kind of contrasts as on plate 7. Note, for example, the range of 0.69 

 foot in the observed elevations in the period July 17-25, 1910. In the same 

 period the total range of the corrected elevations was only 0.21 foot. Note, 

 also, the extremely high points on the dot-and-dash graph on August 5 and 

 August 25 and the relative smoothness of the continuous graph near those 

 points. 



