194 A NEW METHOD OF ESTIMATING STREAM-FLOW 
from which it is seen that when the mean temperature is 28° F., and there is an 
abundant amount of snow and ice on and in the ground available for melting, there 
is a constant addition to the storage of 0.0397 inch of depth per day. For each 
day for each degree that the mean temperature for the day is less than 28° F., there 
is a subtraction of water from storage due to freezing at the rate of 0.00457 inch of 
depth per day. For each day for each degree that the mean temperature for the 
day is greater than 28° F., there is an addition to storage by melting at the rate of 
0.0519 inch of depth per day. The rate of addition to storage by melting is there- 
fore about ll times greater than the rate of subtraction therefrom by freezing 
V0.457 / 
The critical temperature 7 7 " = 28° F. is less than 32° F., the freezing point, 
by somewhat less than the daily range of temperature, because a rapid melting 
during a few hours must be offset by slow freezing during many night hours if the 
net melting is to be zero for the 24 hours. 
Subsequent to the derivation of the above values, it was interesting to find that 
a value for M had been obtained by Mr. Robert E. Horton (The Melting of Snow, 
Weather Review, vol. 43, 599) by an entirely different method. Horton melted 
snow in the laboratory and measured the rate of accumulation of melted snow at 
the base of the cylinder. He concludes that the rate of melting is from 0.04 to 0.06 
inch per 24-hours per degree F. Incidentally, the fact that Horton's value for the 
rate of melting — obtained by direct observation — agrees well with the value ob- 
tained by the methods used in this investigation, tends to vindicate the use of 
assumption No. (10), especially, since upon the validity of that assumption, more 
than any other, rests the determinability of M by this method. 
From the nature of the derivation of equation (92), and because of the units in 
which it is expressed, it should apply to any watershed anywhere in the world, 
hence in applying this method to other streams, the labor involved in deriving 
equation (92) probably need not be repeated. 
SUMMARY OF CONSTANTS DERIVED IN STREAM-FLOW INVESTIGATION 
For convenience in reference for further discussion, all of the constants derived 
on the stream-flow investigation on Streams A and B up to this point are summar- 
ized in Table 51. These represent all the constants derived in the stream-flow 
study except the frequency-ratio constants shown in the last column of Table 66, 
page 218. 
ACCURACY OF COMPUTED NORMAL STREAM-FLOW 
In Table 51 are shown the various normal-flow constants S c , R\, R\, R'i . . . 
R' l0 for Streams A and B, both in absolute amount and converted to percentages, 
with the probable errors. These probable errors are a measure of the accuracy 
which is the best that can be obtained, provided the errors in the derived constants 
are all accidental in character. Assuming that the errors are all accidental, that 
is, that there are no systematic or constant errors affecting the results, it is an even 
chance that the actual error in any constant is greater than or less than its computed 
probable error as shown. For example, on Stream A, R\ = 0.0087, expressed in 
per cent, and its probable error is ± 0.0022. This means that the chances are even 
for and against the proposition that the true value of R\o for Stream A lies within 
