elevations of more than 1,000 or 1,500 feet above sea level. This chart furnishes a conclusive answer to those who believe that floods, except, of course, torrents in the mountain creeks, are caused by the precipitation on the comparatively small area of the watersheds at the headwaters of rivers. If it be granted that forests control the flow of streams, and I doubt that they do except as stated above, it will be necessary, in order to have an appreciable effect on navigable or other important rivers, to reforest areas many times in excess of anything that so far have been contemplated. The rugged mountain slopes and tops, where land has little value, are unimportant as flood producers. It will be necessary actually to reforest the lower slopes and valleys where the land is of great value and where it should be devoted to agricultural purposes. I can not escape this

conclusion.

ARE FLOODS INCREASING!

Two papers have recently appeared, in both of which the argument is made that there is a marked tendency toward increasing flood frequency as a result of deforestation. The first of these papers in point of time was that of Mr. M. O. Leighton, Chief Hydrographer United States Geological Survey." The second paper appeared in volume 2, Senate Document No. 676, beginning at page 112, and later as Forest Service Circular No. 176, January 11, 1910, under the signatures of Mr. William L. Hall, Assistant Forester, and Mr. Hu Maxwell, expert.

Before entering upon a discussion of these papers I wish to draw attention to the following statement in the last-named paper, page 3: Both the Geological Survey and the Forest Service have secured data,b and the results warrant the statement that unmistakably floods are steadily on the increase in some of our most important rivers.

* * *

I wish to remark in connection with this statement that substantially all of the data used by the authors of the papers above mentioned were drawn from the records of the United States Weather Bureau.c

In Water-Supply Paper No. 234 the author has made a diagrammatic arrangement of data composed of annual and decennial means, whereby he shows an apparent progressive increase in the number of flood days at Wheeling, W. Va., and other points, without a proportionate increase in the amount of precipitation. It appears to me that his argument is defective in at least two particulars.

First. The flood or danger stage of the rivers at the various places discussed by him was long ago fixed by the Weather Bureau as being at the point where the river either overflows its banks or damages property adjacent thereto. Mr. Leighton has disregarded these points and arbitrarily assumed, for the purpose of his discussion, a

a Report of National Conservation Commission, p. 95, and Water-Supply Paper No. 234.

The italics are mine. (Author.)

cIn Water-Supply Paper No. 234 the impression seems to be given that the author's researches and conclusions are based on a consideration of river discharge measurements. It is proper to state that river discharge measurements were begun under the direction of the United States Geological Survey in 1896 and that gauge readings of heights of rivers were begun by the United States Signal Service, now Weather Bureau, in 1874.

considerably lower stage in each case, so that his argument fails completely so far as it relates to flood frequency; for example, at Wheeling, W. Va., he assumes a stage of 20 feet, whereas the Ohio at that point is not in flood until a stage of 36 feet is reached. What the author is discussing is therefore not floods as such, but moderate stages of the river.

What appears to me to be a second defect in the author's argument lies in his acceptance of the total number of so-called flood days (20 feet or more being a day of flood) divided by the annual precipitation as an indication of flood intensity, since the annual rainfall, as he himself acknowledges," bears little or no relation to floods. Greater floods may occur during a year of deficient precipitation than during one of excessive annual precipitation if the proper proportion of the rainfall be concentrated over a limited area în a limited time.

An examination of the data for Chattanooga, Tenn., given by that author, discloses the fact that there has not been any increase in the number of so-called flood days at that place. The average amount of precipitation for each daily river stage of 20 feet or more, as determined by him, is almost exactly the same for the two periods, 1884 to 1895 and 1896 to 1907, inclusive. But in the number of actual flood days, as determined by Professor Frankenfield, the official in charge of the Weather Bureau river and flood service, that is, 33 feet or over (and the river does not reach the danger or flood stage until it stands 33 feet above low water), there was a considerable decrease in the second period, in harmony with the precipitation.

I understand that when Mr. Leighton speaks of "the ratio of the annual number of days of flood to annual precipitation," he means the number of days (stage above 20 feet) in each year divided by the total precipitation for the year. Thus, if the number of flood days in any one year is 20, and the total precipitation is 40 inches, the ratio would be 20 divided by 40, or 0.5. These ratios are totaled in eleven-year periods and the average of each period obtained. The average for the first eleven years, as obtained by him, was 0.38, and of the second, 0.48, indicating, in his opinion, an increase in flood intensity during the second eleven-year period, as 1 inch of rain made only 0.38 of a flood during the first period, while in the second period 1 inch of rain made 0.48 of a flood. In other words, during the first eleven-year period 1 inch of rain made only 38 per cent of 20 feet of water, or 7.6 feet; while during the second period 1 inch of rain made 48 per cent of 20 feet of water, or 9.6 feet.

This line of reasoning leads to wrong conclusions, as it is certain that the ratios obtained by dividing the number of days that a certain gage reading was reached or maintained by the annual, or for that matter by any other, precipitation, without entering into the problem the exact height of water gives a meaningless result. It appears to me to be a fatal method of reasoning to take simply the number of days that a stage of 20 feet was reached, without regard to heights above 20 feet. Therefore, if on a certain number of days the gage reading was exactly 20 feet, one would get precisely the same quotient as he would if on the same number of days the gage readings were largely in excess of 20 feet.

a Page 22, Water-Supply Paper No. 234.

I now come to that part of the discussion in Water-Supply Paper No. 234 which has been widely quoted by the adherents of the forestcontrol idea, viz, the proposition that, although the flood periods in the Tennessee have decreased in later years due to diminished precipitation, the flood tendencies have increased. This idea, like others that have been put forward in this connection, is important if it can be substantiated; and if it is proven, then it is incumbent upon the author of that paper to show that the increase is due to deforestation, which he does not do. I have no data as to the area. that has been cleared or that has been allowed to revert to forests, but it can not be great in twelve years.

This whole matter of the influence of forests upon climate and floods is so important to the nation in planning a correct economic policy for the future that we should move cautiously and be sure that we are building safely and wisely. I am heart and soul with the noble men and women who, as individuals or collectively, are striving to protect and conserve in the interests of the whole people the nation's resources of forest and field and of minerals and water power, and in this opinion I am generally and strongly sustained by the scientific staff of the Weather Bureau.

With regard to the matters of which this paper specifically treats I wish for the freest, fullest, and fairest discussion and investigation, with the end in view of correcting error if there be such and of finding common ground upon which all well-meaning persons may stand. Those whose official reports differ from mine I believe to be as honest and as sincere in their investigations and conclusions as I know myself to be.

To return to Supply Paper No. 234, referred to above, I quote as follows from page 23:

*

The results for the Tennessee basin cover twenty-four years, from 1884 to 1907, inclusive. * * Summing up the flood-producing rains for the twenty-four year period, it is found that the total is 335, of which 313 occurred from December to May, inclusive, and the remaining 22 during the other portion of the year. It is apparent that the number of such rains from June to November is not sufficient to afford a basis of comparison. Therefore only the December to May floods will be considered.*** On dividing the period covered by these 313 floods equally, two consecutive twelveyear periods are afforded, which give a basis of comparison. The floods in the later period, resulting from a given depth of storm precipitation, are clearly shown to be more severe than in the earlier period. The method of presentation further makes it possible to compute the increase in flood tendency due to deforestation in the Ten

nessee.

*

If we now divide the number of flood days by the number of storms the result will be the number of days per storm.

The algebraic sum of the above percentages is 149 and the average is 18.75, sums up the effects of deforestation on run-off from 1884 to 1907, inclusive.

which

I invite attention to the figures given in this table "Days of flood per storm." If the run-off in the second period was greater than in the first, due to deforestation, would not the latter show a uniform and progressive influence increasing as the amount of rainfall increased? How, then, does it happen that a decrease in run-off of 43 per cent is shown for rains of intensity 1 inch to 1.5 inches, while in the next higher grade of intensity, viz, 1.5 to 2 inches, an increase of 80 per cent is shown? In the next higher grade, viz, 2 to 2.5 inches, the increase drops to 4 per cent. These results founded, in my judgment, on incorrect premises, are both inconsistent and meaningless. To "divide the number of flood days by the number of storms" gives no valuable quotient, for the gage readings selected as floods are not floods, but only moderate stages, and no account is taken of the actual height of the water, and while the conclusion is reached that there is an increase in flood intensity of 18.75 per cent in the Tennessee basin in the past twelve years due to deforestation, no records or other evidence are presented that there is not as much forest area in this basin as there was twelve years ago; or that, if there is a decrease, it would be sufficient to account for such a large increase in flood intensity.

But-and here is the most important matter in the consideration of Mr. Leighton's conclusions-no matter how complete the data may be, or how fundamentally sound and fair its collation and grouping, the comparison, the one with the other, of such short periods as those measured by only twelve years, can not give results with regard to changes in climate and floods that will permit the most skilled meteorologist or engineer to draw fundamental conclusions that can have any value. Precisely the same amount of rain falling in the two periods and no change whatever in forest or cultivated area might produce largely differing results on floods, depending on the sequence with which it fell over the different tributaries and how it was concentrated or scattered, and on many other complicated conditions of run-off, such as the coinciding of the flood volume from one tributary with that of another, instead of each passing down the main stream at different times.

There is also the difficulty of securing accurate precipitation data. Whenever the height of the gage is altered or other change made in its environment that disturbs the flow of the air currents the readings of one period may not fairly be compared the one with the other. These defects vitiate the precipitation data of many stations of the Weather Bureau, especially those in large and growing cities, and can only be remedied by the Government controlling for a long period of years an area at each station so large that it can determine the exposure and keep it constant.

Another way of comparing the precipitation and the river stages of the Tennessee basins. I give in the following table the rainfall at Chattanooga and Knoxville separately for the months December to May, inclusive, for each of the twenty-four years considered in Water-Supply Paper No. 234; also the total number of days of river stages of 20 feet and above on the Chattanooga gage. The rainfall so tabulated includes only the heavy rains, and the arrangement according to intensity is precisely the same as that followed in WaterSupply Paper No. 234.

Heavy rains at Chattanooga and Knoxville, Tenn., during six months of each year (December, 1883, to May, 1907).

The data of the above table have been divided into two periods of twelve years each, with the following results:

Total amount of the above heavy rains as per table, 404.61 inches. Dividing this total by two, to get the approximate average of the heavy rains for the watershed, we get 202.30 inches.

Total number of days with stages of 20 feet or more at Chattanooga.....

166

Dividing the amount of the heavy rains in the watershed by the number of days 202.30 with a river stage of 20 feet or over, we get 166

=1.220 inches as the amount of rain

fall that probably produced one day of a stage of water in the river of 20 feet or more.