112 EFFECTS OF WINDS AND OF 



be. The computed probable errors are therefore somewhat too large to rep- 

 resent the true degree of accuracy of a single corrected elevation. 



From various sources of information, some of them external to this in- 

 vestigation, the following estimates of the fluctuations of mean lake surface 

 of Lake Erie have been made. The estimates are approximately applicable 

 to Lake Michigan-Huron, with the one exception that the net outflow from 

 Lake Michigan-Huron is smaller, in the units here used, than from Lake 

 Erie and less variable. 



It is estimated for Lake Erie that during the months June to October of 

 each year the net outflow, namely, the outflow through the Niagara River 

 minus the inflow through the Detroit River, corresponds to a fall in the mean 

 lake surface varying from +0.016 foot on some days to 0.002 foot at the 

 other extreme for some days. Similarly, it is estimated that the rainfall on 

 the lake surface causes a rise in the mean lake surface varying from zero on 

 days of no rain to +0.085 on rare days of very heavy rainfall; that the run- 

 off into the lake from the surrounding land corresponds to a rise varying 

 from +0.004 foot on some days to +0.040 on other days; and that the 

 evaporation corresponds to a fall in the mean lake surface of from 0.000 

 foot on some days to +0.021 foot on days of extremely rapid evapora- 

 tion. 



For the season of 1910, the five months June to October, inclusive, the 

 algebraic sum of these four influences was a fall of about 0.004 foot per day 

 for the whole season. 



An inspection of the monthly means of corrected elevations as shown in 

 table No. 26 makes it clear that in general during the season June to October 

 the mean lake surface is falling at a mean rate usually between 0.00 and 

 0.02 foot per day. 



It is clear from the three paragraphs next preceding this that the actual 

 variation of the mean elevation of the whole surface of any one of the Great 

 Lakes is, as a rule, as much as 0.01 foot in two days, that it is frequently 

 more than 0.02 foot in 24 hours, and that on rare occasions it may exceed 

 0.08 in that period. 



This estimated rate of variation was taken into account in fixing the size 

 of the group to be used in taking each mean in such tables as Nos. 19 to 24. 

 If the group were taken very small, as, for example, two or three days only, 

 the mean would be unstable, and the two or three residuals would give a very 

 poor computed value for the probable error. On the other hand, if the group 

 were made large, to include, say, eleven values, then the extreme residuals in 

 each group would depend very largely upon the actual variations in the mean 

 lake surface commented upon in the four preceding paragraphs, and so would 

 not be of value as indicators of the accuracy of the corrected elevation for 

 each day. In an eleven-day group, the extreme residual at each end of the 

 group would include the variation in elevation of the mean lake surface for 

 five days, the interval between that residual and the middle date of the group. 

 For the mean for the group most nearly represents the middle date in 



