survey of the forests now in progress indicates that the period of maximum deforestation was about 1850.

In order to have a safer basis for conclusions as to streamflow than would be furnished by my four years' series of stream gaugings in New Jersey, all the long series measurements of flow available for streams of New England and the Middle States were collected and studied. My conclusions, consequently, apply to these water-sheds as well as New Jersey.

The following series of measurements were thus used: Sudbury river, Massachusetts, 16 years long; Connecticut river, 8 years; Croton river, N. Y., 14 years; Passaic, N. J., 17 years; Tohickon, Neshaminy and Perkiomen creeks, Pa., each 7 years; Potomac river, 6 years. These streams have from 7 to 53 per cent. of their catchments in forest. In New Jersey, in addition to the Passaic series, I was furnished a series 8 years long on the Hackensack, and measured the flow of a dozen other streams for periods of 4 years or less. These have from 13 to 88 per cent. of their catchments forested, so that the range was wide enough for safe conclusions. A difficulty appeared, however, in the fact that the percentage of forest usually varied with the surface geology, and the latter affects stream-flow to a greater extent than forests. Taking evaporation to mean the difference between total rain-fall and total run-off of the streams, the important fact was developed by these studies that the amount of rain evaporated is never directly proportional to the total rain-fall, as is often assumed to be the case. Evaporation is increased slightly for increased rain-fall, but rapidly with increased temperature of the atmosphere. Stream-flow was found to be mainly a function of rain-fall and temperature, and little or no effect upon the total yearly run-off was traceable to forests or other vegetation, or to topography. For instance, the Sudbury catchment, with 14 per cent. of forest; the Croton, with 30 per cent., and the Passaic, with 44 per cent., all having about the same mean temperature and similar topography, show the same total flow-off for a given annual rain-fall. The vegetation on the unforested portions of these catchments is much the same in each case. We find here no effect upon evaporation, from forests. The Connecticut, with 53 per cent. of forests and temperature 2.7 degrees lower, shows much less evaporation and larger flow than the above streams, but the

Potomac, with about the same forest as the Connecticut and higher mean temperature, shows much more evaporation, the latter in each case varying with the mean temperature without regard to the amount of forest. Neshaminy creek, in Pennsylvania, having 7 per cent. of its area in forest and the remainder under a high state of cultivation, shows 10 per cent. more evaporation than the Sudbury, which can only be accounted for by temperature. The Passaic and Raritan, in New Jersey, each drain a little over 800 square miles. The Passaic has 44 per cent. in forest, the Raritan 13 per cent., or nearly the same as the Sudbury. It shows 4.3 per cent. more evaporation than either the Passaic or Sudbury. The Hackensack, like the Raritan, lies on the red sandstone plain, but has 60 per cent. of forest, nearly five times as much as the Raritan, but it shows about the same evaporation-about 4 per cent. more than the Passaic, with 44 per cent., or the Sudbury, with only 14 per cent. of forest. This is exactly accounted for by the temperature. Again, the Great Egg Harbor and Batsto, streams of Southern New Jersey, have 88 per cent. of their catchments forested with pine on upland and cedar in swamps, but this large proportion of forests and small area cultivated does not prevent an increase of 14 per cent. in evaporation over that of the Sudbury, Croton and Passaic, or of 10 per cent. over the highly-cultivated Raritan catchment. The increase of evaporation is again accounted for by increased temperature and uninfluenced by forests.

I know of no more accurate way to compare the relative evaporation from forested and deforested areas, than by measuring the rain-fall and the amount of rain flowing off, provided the observations are long enough continued and begin and end with full ground-water. In this way we obtain natural conditions, and include both direct evaporation from the soil and also the water drawn up by vegetation, much of which is exhaled into the atmosphere. The loss into the earth which does not re-appear in the natural drainage channels, may, under ordinary circumstances, be neglected. This method is certainly far preferable to the attempts to measure evaporation on a small experimental scale, which have frequently been made, and the results of which are often quoted. Some of these measurements entirely neglect the large draughts of moisture made

by the trees of the forest. All of them fail to obtain the same conditions which prevail naturally over large areas of forests. I am, consequently, forced to conclude, from my studies, that the effect of our New England and Middle States forests upon the total flow-off of streams, hence upon evaporation, is not important enough to be shown in the measurements of streamflow.

A further analysis of observations on the Passaic catchment indicated that while the total average annual evaporation for an annual rain-fall of 45.00 inches was 22.70 inches, the portion of this taken up by vegetation alone was but 6.03 inches, or 27 per cent. of the whole, indicating at once the minor importance of vegetation. The powerful influence of the temperature of the air upon evaporation may be appreciated when we recall the well-known fact that the moisture which may be held in the air is doubled for each increase of 20 per cent. in temperature. Again, Mr. Desmond Fitzgerald made observations from 1875 to 1890, at Boston, which indicated an annual average evaporation from water surface of 39.20 inches, whereas, observations of rain-fall and stream-flow on Sudbury water-shed for the same years show an average evaporation from the earth's surface of 23.14 inches only. The air was, therefore, able to take up 16.06 inches more moisture from the water than it obtained from the land; hence, it is to be considered the powerful factor in determining evaporation. Again, in support of our conclusions, if evaporation is so much less from forested than from cultivated areas, as is sometimes claimed, should not forests be found thriving in full variety and luxuriance far beyond the limit where rain-fall becomes too light to support other vegetation?

Next we have to consider the effect of forests upon the greatest and least flow of streams: A careful investigation of recorded greatest and least flows per square mile of catchment, yields no results, indicative of any important effect due to forests. The maximum freshet usually occurs either when the ground is frozen and a warm rain comes on a heavy covering of accumulated snow, or else in summer, when an unusually heavy rain falls upon ground already saturated. The rate of flow-off is then mainly determined by the topography of the catchment. So the lowest flow occurs when the stream has for a long time been drawing upon stored ground-waters, and has

drawn down such waters to a point far below where they can be influenced by any surface conditions. The rate of flow is then affected mainly by the capacity of the earth of the catchment for holding water and its rate of yielding up the same. matter not of vegetation, but of surface geology; therefore forests can have little effect upon either the greatest or the least rate of flow of a stream.

Thus far our results are negative, but very beneficial effects were nevertheless observed from a good covering of forest upon the catchment. It became evident from these studies of stream-flow that streams were often supplied for many months entirely from water stored in the ground. For seven months in 1880 and eight months in 1881, the Passaic river was thus supplied. With only rain enough to make good the evaporation our New Jersey country will yield up in nine months ground or spring water equal to from 2.29 to 7.59 inches of rainfall. Small, barren, red sandstone catchments yield the least water, and the sand and gravel of the Tertiary formation the most. If the rain falls uniformly from two to two and one-half inches per month may be taken into the ground, and discharged thence to the streams. The entire rain-fall of an average year, less the evaporation, could be thus taken into the earth, and none need flow over the surface to the stream. Anything which affects the capacity of the earth to take up water, and the rate of discharge of the same to the stream, therefore, affects the streamflow by making it more or less uniform through the year. the surface is forested, a mass of absorbent humus is present; the tree roots, fallen trunks and mosses all obstruct the flow of water over the surface, and hold it until the earth can take it up. The roots increase the friability and absorbent power of the earth; the water is held and absorbed in large quantities, and slowly fed out to the streams.

If

Cultivation also, like forests, increases the absorbent power of the earth, and well-cultivated areas will, I find, take up as much water as forested areas. Here, however, drains are provided and water courses opened to hasten drainage of the ground-water into the streams. As a result, the discharge of streams after a rain is quicker, freshets are more frequent, the dry periods longer. This is admirably illustrated in the case of the Passaic and Raritan. The former has 44 per cent. of its

catchment forested; the latter only 13 per cent., the balance being well drained and highly cultivated. Each will draw about four inches of water from the ground in nine months, but during the first three months the Raritan discharges 2.52 inches of this, the Passaic 2.10 inches, or 17 per cent. less. During the last three months, however, the Raritan discharges .70, the Passaic .87 inches, or 25 per cent. more. The Passaic, consequently, has a much more equable flow. During four years, freshets exceeding 10 cubic feet per second per square mile, occurred but five times on the Passaic, against thirteen times on the Raritan, whereas the Raritan sank below 2 cubic feet per second per square mile, an average of 271 days each year; the Passaic only 245 days. While in the case of these particular streams this result may be due partly to surface geology, the same facts were observed throughout the list of measured streams. Those having the largest proportions of forest upon the catchment invariably show the best sustained dry season flow, although the total run-off is no greater.

Our Southern New Jersey streams have the most forest, and are remarkably steady, the dry season flow averaging double that of the northern rivers. This is undoubtedly due in large part to the great absorbent power of their sandy catchments, but a critical study of their daily flow reveals the important contribution of the cedar swamps to this result. Should these be cut off the streams would suffer, becoming much more unreliable.

It will be seen that as between cultivated and forested catchments, therefore, our gaugings indicate the same total run-off for a given rain-fall, but a much more uniform discharge, fewer floods and shorter dry periods on the forested areas. Forested streams are consequently more valuable and reliable for water power, and for water-supply they require less storage reservoir capacity to provide for a given daily draught. The waters are also much less likely to become muddy or otherwise contaminated.

The worst condition of all for a catchment is barrenness. Barren earth is non-absorbent; the water fails to penetrate it and oxidize its fertilizing constituents. Heavy rains run over the surface, washing off all of the loose material, and barren conditions once inaugurated perpetuate themselves. There is