228 A NEW METHOD OF ESTIMATING STREAM-FLOW 
Stream B, the estimate of power available from the computed mean duration curve 
(Plate 19) would have been -practically identical with that made from the observed 
mean duration curve. For estimates for plants utilizing storage, the daily variation 
in flow is not so important as the monthly and annual aggregate flows. In this 
connection note the agreements on Stream B (Tables Nos. 57, 59a, pages 206 and 
207 and footnote on page 230.) 
The above rules for applying the method are written in definite form from 
which it might be inferred that the results will be unfailing and uniform. It is not 
claimed that the new method of computing the constants of a stream and of esti- 
mating the future stream-flow is standardized so far that it may be applied by 
following a set of rules. It is still in the developing stage. The method and the 
computation of the constants involves the use of judgment, as distinguished from 
rules, at many points. The computations involve essentially a process of successive 
approximations which may be very slow and poor if poor judgment and little 
knowledge is used. 
APPLICATION TO FORECASTING THE FLOW OF A STREAM 
There are various cases occurring in engineering practice after a plant utilizing 
stream-flow is in operation, in which it is desirable to make and act upon a predic- 
tion of the stream-flow a few days or weeks in advance. Contracts for the sale of 
power may thus be made more wisely. Coal may be bought to better advantage 
in the case of interconnected systems involving both steam and water-power. The 
new method of estimating stream-flow lends itself admirably to such predictions. 
Notice that S c and R'i<>ri are known four months ahead. S c , R\r s , R\r, and 
R'loTia are known a month ahead, and all terms even of the flood-flow formula 
having subscripts greater than 5 are known nine days ahead. Predictions four 
months ahead, one month ahead, or nine days ahead of time would be made by 
using the known terms of the formula and supplementing them by average values of 
the remaining terms for the corresponding portions of the year, based upon past 
experience. 
Since the energy output of a turbine varies as the three-halves power of the 
head, a small gain in head is of relatively more importance in a low-head than in a 
high-head plant. The prediction method stated in the preceding paragraph will be 
especially useful on the low-head plant, in rendering possible the maintenance of a 
maximum head and at the same time wasting as little water as possible over the 
spillway. 
In the case of a high-head plant utilizing storage, it is probably more important 
to know how much water a given rain will put into the reservoir, or how much 
water will run into it as the result of a thaw in the spring, than to forecast the flow 
from day to day, inasmuch as the daily flow may change the reservoir content but 
little. This method lends itself admirably to forecasts of this kind also. Take the 
case of the rain of October 4-5, 1911 (see plate 13). On October 5, the total 
amount of water which the rains of October 4 and 5 would deliver to Stream B by 
surface travel could have been forecasted by equation (86), page 177, with an 
error of less than 8 per cent. Inasmuch as the flow by percolation was relatively 
insignificant in this case, the estimated total flow during the month following 
October 5 due to this rain would have been but slightly in excess of that estimated 
by equation (86) alone. If the temperature should rise suddenly in the spring, 
when the ground is covered with snow, the amount of water (melted snow) to be 
