BAROMETRIC PRESSURES ON THE GREAT LAKES 117 



land, analogous to table No. 29, showed clearly a short-period seiche, but 

 furnished no clear evidence of a long-period seiche. There were 252 periods 

 in the table. Of these 252 periods, 165 were not less than 2 hours nor more 

 than 3.5 hours. The mean of these 165 periods was 2.6 hours, which is 

 believed to be the prevailing seiche at Cleveland. 



A study of the Cleveland graphs indicated that a long-period seiche might 

 be in existence there much of the time, and yet little or no evidence of 

 it would appear in a frequency-distribution table similar to table No. 29, 

 because the short-period seiche (2.6 hours) was rather persistent and be- 

 cause its shortness of period tended to produce well-defined maxima and 

 minima, which would tend to conceal in such a table any long-period seiche. 

 Accordingly, three determinations of a long seiche period at Cleveland were 

 made by the method of selected principal maxima and minima. The 

 selection was made by somewhat arbitrary rules based on the general prop- 

 osition that the purpose of the selection was to remove the mask made up 

 of short-period oscillations, mainly 2.6 hour seiches. The first determination 

 was based on one selection of a moderate number of principal maxima and 

 minima and a frequency-distribution table. The second and third determi- 

 nations were based on a second selection of principal maxima and minima. 

 The three determinations involved, respectively, 29, 31, and 30 identified 

 periods. The three mean periods from the three determinations were in 

 order 13.9 hours, 12.5 hours, and 12.9 hours. The mean of these is 13.1 hours, 

 which is believed to be the period of the prevailing long seiche at Cleveland. 



The determination of this period, 13.1 hours, at Cleveland, as indicated 

 above, is weak. 



But there are three supporting lines of evidence indicated below, which 

 greatly strengthen the evidence indicated above in favor of a 13.1 period at 

 Cleveland. 



First, it is to be noted that the mean, 13.1 hours, of three widely separated 

 values, 13.9, 12.5, and 12.9, agrees exactly with the 13.1 hour period de- 

 termined independently at Buffalo on the same lake. 



Secondly, there were 6 days of the 22 examined at Buffalo and Cleve- 

 land which were common to these two stations. On these 6 days the maxima 

 at Cleveland of the supposed 13.1 hour seiche coincided in time, within the 

 errors of identification at the two stations, with the minima at Buffalo of the 

 supposed 13.1 hour seiche there. So, also, the minima at Cleveland co- 

 incided closely with the maxima at Buffalo. 



Thirdly, the theoretical period of oscillation of Lake Erie lengthwise was 

 computed and found to be 13.1 hours. This, combined with the second line 

 of evidence, clearly identifies the Cleveland and Buffalo 13.1 hour seiches 

 as one and the same thing, as the theory shows that in that case the relation 

 of maxima and minima at the two stations should be that which was actually 

 found. 



The theoretical period of oscillation of Lake Erie lengthwise was found by 

 use of the table on page 618 of the Coast and Geodetic Survey Report for 



