In determining the quantity of water that may be taken from a small stream the area of the watershed answers the same purpose as the area of the roof which delivers water into a tank, the only difference being that from the roof all the water is always delivered, except a small proportion that evaporates at the beginning of a rain in summer. From the surface of a watershed, on the contrary, a large amount, and in some cases all of a stream, will be absorbed by the ground and by the vegetation and will never be delivered into the stream which drains an area. On large streams it is fair to assume that, on the average, only one half of the rainfall on the area will reach the stream, while with sandy soils this may be as small as 20 per cent. From December to May inclusive, when the ground is frozen, when there is no vegetation to absorb the water, and when evaporation is very light, practically all of the rainfall reaches the streams. From June to August, on the other hand, when the soil becomes rapidly parched, when vegetation is most active, and when evaporation is high, frequently no rainfall reaches the streams and the ground water sinks lower and lower, so that often streams themselves dry up. It is necessary, therefore, in providing for a definite quantity of water to be taken from a reservoir built on a small stream, to make the reservoir large enough to furnish water from June to September without being supplied with rain. This does not call for a very large dam or a very large storage, and three months' supply will usually be ample. We have already estimated above that the quantity of water needed for three months will be 45,000 gallons, or about 6000 cubic feet. If the reservoir is built in a small gulley or ravine, its width may be twenty-five feet. If the length of the reservoir or pond formed by the dam is 240 feet, then the reservoir will furnish 6000 cubic feet for every foot of depth, and a reservoir of that size holding one foot of water will tide over a dry season. Evaporation during these same three months will use up about a foot and a half in depth over whatever area the reservoir covers, so that two and a half feet in depth must be provided above the lowest point to which it is desirable to draw off the water. It would be well to allow a depth of at least ten feet in order to avoid shallow, stagnant pools, and if this depth is provided, even more than the two-and-a-half foot depth mentioned might be withdrawn in extremely dry seasons, though perhaps at some reduction in the quality of the water. _Deficiency from well supplies._