182 A NEW METHOD OF ESTIMATING STREAM-FLOW 
The values in parentheses are values of (n — G), the illustration of the computa- 
tion of which has yet to be given. The unit is 0.01 inch of depth to (n 4 — G t ) inclu- 
sive, and 0.1 inch of depth thereafter. The coefficients of these quantities are the 
R' /s of equation (87). The sum of the products, R' f (n — G), which are in 0.001 
c.f.s., gives D f . D t added to D„, computed on page 180 and their sum added to 
—D' gives —D. The various quantities can be identified by the column headings. 
The above computation of D f and — D is given as if beginning on March 31. 
Earlier than March 31 the mean air temperature on the watershed does not rise 
high enough to produce a large enough amount of melting to create a flood-flow, 
hence the R' /s and the (?'s do not enter into the computation of — D before that 
date. Preceding March 31, — D is merely the sum of D„, computed on page 180 
and — D' ; that is, preceding March 31, D/ is zero because there were no flood-flows 
due either to melting or to rain. The computation of — D for the dates preceding 
March 31, which will appear later in the example of observation equations for those 
dates, page 191, can be verified by adding the D„'s of page 180 to —D' scaled 
from Plate 10. 
ESTIMATING DATES TO USE IN COMPUTATION OF D, AND OF (n-G) 
Only certain dates enter in the computation of D f , as shown in the particular 
illustration. These are dates on which it is believed that there was sufficient melt- 
ing to create a flood-flow. In order to decide which dates to use, it is necessary to 
assume values of C, M and T" , and compute n for the stretch of dates near the 
beginning of the period of rapid melting. Whenever these estimated values of r x 
exceed G, the date is included in the computation of D f . 
Since Solution AA on Stream A was made after Solution M, the values of C, 
M and T" used in Solution M could be used in estimating the dates to use in the 
computation of D f . There were two modifications made in these values, however, 
the necessity for which became apparent after the completion of Solution M and its 
successors on Stream A up to Solution A A, and after completing Solution X on 
Stream B. The first modification consisted in reducing M from +8.39 to +5.00. 
The evidence in favor of a reduction in the size of that M was clear, and was of the 
following kinds: The values given in (62) were used in the computation of the 
stream-flow for Stream A which is shown on Plates 9 to 12, inclusive. In a study 
of these graphs, in connection with the computations in the least-square solutions, 
clear evidence was found that M= +8.39 is too large. Such evidence was of three 
kinds, as follows: (a) the rate of melting near the close of the season — during the 
supposed period of decrease of melting — as computed appeared to be too rapid to 
be plausible. In some cases it appeared to wipe out the whole of the remaining 
snow in a single day, whereas it was known to last a week or more; (6), the observed 
curve of discharge is much less sensitive to winter thaws and early spring thaws 
than the computed curve of discharge based upon M = +8.39. This can be verified 
from Plates 9 to 12, inclusive; (c), in the derivation of the above value of M the 
flood-flows were not taken into account by the flood coefficients, R' t . On days of 
flood-flow the M was forced to do double duty, express the rate of melting and also 
the flood-flow. Hence its derived value came out too large. 
It may be worth while to discuss the evidence mentioned in (6) more in detail. 
Attention is directed to the computed and observed stream-flow for the period 
November to February of each winter of 1912-13 to 1914-15 shown plotted on 
Plates 9 to 12 inclusive. On November 5 to 9 and 19 to 22, 1915 (Plate 12) the 
