A NEW METHOD OF ESTIMATING STREAM-FI." >\Y 
205 
errors which are not small in comparison with the error introduced by the assump- 
tion of a constant — . It is to be noted in this connection that (I) the distribution 
E„ 
of precipitation on these areas is fairly uniform throughout the year; and (II), that 
the vegetation is of a kind which does not change its foliage considerably from 
season to season. That is, most of the trees are evergreens and, except for the 
grass, the change in loss of water from the ground supply from season to season 
due to absorption by the growing vegetation would presumably be small. Both 
these conditions would favor a constant — . On the other hand, a watershed 
E„ 
subject to large variations in the annual rainfall and large variations in vegetal 
cover from season to season due to change in foliage and crop harvesting might 
require a variable — . 
E w 
ACCURACY AS TESTED BY DETAILED NUMERICAL COMPARISON— COMPARISON OF DAILY 
OBSERVED AND COMPUTED FLOWS 
The daily numerical values of the computed stream-flow have been compared 
in detail with the daily observed stream-flow for the whole 58 months, March 1, 
1911, to December 31, 1915, on both streams. This comparison is shown in Table 
54. 
Table 54 — Comparison of daily observed and computed flows, all days 
Percentage of days on which computed flow agrees with 
observed 
Within 10 per cent 
Within 20 per cent 
Within 50 per cent 
Stream A 
Stream B 
27 
28 
54 
62 
87 
97 
COMPARISON OF FLOOD WITH NON-FLOOD MONTHS 
Even a casual examination shows that the agreement of computed with ob- 
served stream flow is much better in non-flood months than in months in which 
Table 55 — Comparison of daily computed and observed flows in non-flood against flood months 
Percentages of days on which computed flow agrees with 
observed 
Within 10 per cent 
Within 20 per cent 
Within 50 per cent 
Stream A 
Non-flood months . . . 
Flood months 
Stream B 
Non-flood months . . . 
37 
17 
46 
24 
67 
30 
81 
43 
99 
72 
100 
85 
there is flood-flow. Using the criterion of flood-flow already explained, each month 
in which there is a day of computed flood-flow as great as 0.030 cubic foot per second 
