18 BULLETIN 854, U. S. DEPARTMENT OF AGRICULTURE. 
For determining the velocity, coloring matter can be used success- 
fully only in clear water. Potassium permanganate as well as dyes 
of all colors were tried in muddy water having a large velocity and 
it was found practically impossible to detect the colors. However, 
the voltmeter method may be used equally well in muddy and in 
clear water for determining the velocity. It is believed that veloc- 
ities obtained by the use of either color or voltmeter will be quite 
accurate if the mean of several readings is taken. 
RESULTS OF OBSERVATIONS. 
Tables 3, 4, and 5 give the results of the tests. The various series 
are arranged in ascending sizes of tile and ascending grades. Table 
3 gives the results of observations on clay tile; Table 4, the results of 
observations on concrete tile; and Table 5, the results of observa- 
tions on clay tile poorly laid. The Kutter coefhcient of roughness, 
Uj given in column 10, was determined from a large diagram specially 
drawn for this investigation. The variation and irregularity of 
the joints of the tile in the lines poorly laid can be seen in Plate 
VIII, figure 2. 
The tests summarized in Tables 3 and 4 were plotted on coordinate 
paper, with velocities as abscissae and depths of flow as ordinates. 
Mean curves were di^awn for each grade through the points repre- 
senting the tests for each size and kind of tile. These curves are 
shown on Plate IX. A study of the curves reveals some interesting 
facts. For the flatter slopes the cui'ves more nearly approximate a 
straight Hne; as the slopes increase the lines become more curved, 
until at the steepest grade there is considerable bulge to the curve. 
The velocity at any depth of flow over haH fuU is sho^n at a glance. 
It will be noted that the velocities at half full and at full are seldom 
the same, as they would be according to the Chezy formula. The 
greatest velocity seems to be approximately at the 0.8 depth. The 
curves were not extended below the haK-full point on account of the 
insufficiency of data. In the largest sizes of tile, where symbols are 
shown but no curves have been drawn, incompleteness of data has 
prevented the development of accurate cuiwes through these points. 
It wiU be seen that v/ith some of the curves the points lie practically 
on the fines, while with other curves some points vary greatly, show- 
ing probable error in the experiments. 
