22 A NEW METHOD OP ESTIMATING STREAM-FLOW 
METHOD OF COMPUTING DAILY BAROMETRIC CORRECTION 
The mean barometric correction for a day at each of the four gages stated may 
be computed from equations (10) to (13) inclusive, by first computing the effect for 
each hour of the day, summing them up, dividing by 24, then changing the sign 
of the result so obtained. To compute, in this manner, the daily barometric cor- 
rection for each of the 1,500-odd days of observation used at each of the four gages 
would be a long, tedious process. 
The method of computation actually used to secure the barometric corrections 
for each day consisted of five steps, as follows: 
(1) The hourly barometric effects were computed by equations (10) to (13) 
inclusive for the first, the last, and a few intermediate days of the long series of 
days under consideration. The daily barometric correction was obtained for each 
of these selected days, by taking the mean of the 24 hourly effects for that day and 
changing the sign. 
(2) The change in barometric correction from day to day through the whole 
series was computed from the appropriate one of the formulas (14) to (17) inclusive. 
(3) The changes from day to day computed in step (2) were applied one by 
one to the computed correction for the first day of the series obtained in step (1) 
to secure the barometric correction for each day in turn up to and including the 
next selected day on which the barometric correction has been computed in step (1). 
(4) At this point, the second selected day, a discrepancy appeared between the 
barometric correction as computed for the day by step (1) and as computed by 
step (3). This discrepancy was distributed proportionally to time between the 
first and the second selected day, and all intermediate values (computed by step 
(2)) corrected, so as to make the discrepancy disappear at the second selected day. 
The values as corrected were adopted as being sufficiently accurate. 
(5) Steps (3) and (4) were taken from the second to the third selected day, 
from the third to the fourth selected day, and so on to the end of the series. 
The rule ordinarily observed in regard to selected days was to place them not 
more than one month apart in any case and not more than 10 days apart over any 
interval through which the discrepancy to be distributed exceeded 0.070 foot. 
The discrepancy in question is due to two causes : (a) to omitted decimal places 
in the computations, and (6) to the fact that the process of computing the hourly 
barometric effects, and thence the daily barometric effects and corrections, by 
step (1) involves a smoothing out of the discrepancies between the value of the 
constants in equations (14) to (17) inclusive as derived from the least-square solu- 
tion, and therefore does not agree exactly with the computations made from formulas 
(14) to (17). The rule stated in the preceding paragraph was adopted, as a result 
of experience, as probably giving the best balance between extreme accuracy, on 
the one hand, and large expenditure of time in computation, on the other hand. 
WIND EFFECTS 
The wind, in blowing across the surface of a lake, piles the water up on the 
leeward shore. The slope of the water surface due to this cause, after a steady 
regime has been established, is downward to windward. The effect, W, of a wind 
in causing the water-surface at a shore gage for an hour to read higher or lower than 
the mean elevation of the whole lake surface for that hour is given by the expression 
W= +0.088 [^ ) S X (18) 
