of a stream for 20 years it is certain that the record for the next 20 years will not be like it. The problem is to visualize with some degree of accuracy the characteristics of a thousand year record of the stream of which only a 20-year segment is at hand; and to lay out works that will reasonably meet requirements based on the thousand year conception. In doing this the record of all other streams in any degree comparable to the one under discussion may well be analyzed and made to contribute to the general conception of stream flow. The mass curve is used to ascertain the storage required for an assumed rate of draft applied to the flow record of a particular stream. It gives the absolute facts with reference to that stream for that period. When numerous mass curve records are calculated the results may be analyzed by statistical methods, and certain normal curves or factors found to be used as a basis for estimates. Care must always be taken that data so derived are not applied in cases where climatic conditions are different. It has been customary, in the past, to estimate runoff and storage, for streams where long records are available, on the basis of the dryest of record; or in cases where only meager records are available, by means of comparisons per square mile of watershed area with similar streams for which long records are available, making due allowance for factors of rainfall and evaporation. Studies in the last fifteen years have developed more advantageous methods. The following rules give an idea of the way in which they are carried on. 1. Start with a record or estimate of the mean flow of the stream. That is the best starting point for all calculations of flow and storage. Do not make comparisons based on so much per square mile. 2. Estimate the water to be made available as a percentage of the mean flow. 3. State storage, actual or proposed, in terms of the mean annual flow. 4. The mean, in all things, is to be first ascertained, and the probable variations from it are to be then considered. Mean flows Mean flows are usually stated as inches of runoff per annum. This form of statement has the advantage of making them directly comparable to the rainfall figures with which they will naturally be compared. To obtain from this the mean flow in United States. gallons, multiply the area in square miles by the inches of runoff and the product by 47,851 to obtain the mean flow in gallons per day, and by 17,380,000 to obtain the mean flow in gallons per annum. Evaporation from the surface of the water in a reservoir must be allowed for, except that when the records represent conditions after reservoir construction no correction is necessary. The recommended procedure is as follows: First. When the mean annual rainfall is equal to or in excess of the mean annual evaporation from water surface. Two corrections are necessary: (a) A reduction in the estimated mean flow equal to the excess of average evaporation from the water surface of the reservoir over average evaporation from an equal land surface. The corrected mean flow is the estimated mean actual flow after the reservoir is built and is to be used in calculation of yield. (b) A reduction in storage capacity that grows out of the unequal distribution of evaporation at different seasons of the year. This may be compared to the drop in level of a lake with no tributary land area during the summer of the dry year, but refilled by fall rains. In a general way the allowance to be made under this heading for reservoirs in northern New England and New York will be 6 inches in depth over the reservoir area, increasing to 8 inches in the latitude of Philadelphia and to 12 inches more on the south Atlantic slope. No data are available for midwest and western. reservoirs under this heading, but in many cases the allowance would be much greater than 12 inches. A rough rule given in the Civil Engineer's Pocketbook is to use for this depth the estimated mean annual evaporation from water surface less two-thirds the mean annual rainfall. The evaporation found under (b) is deducted from the storage capacity, and the net storage is used in calculation of maintainable yield. Second. Where the evaporation from water surface exceeds the mean annual rainfall the method suggested by Cory* is recommended. The calculation of yield is made directly without either of the reductions noted above, and from the computed net yield is deducted the estimated average evaporation from the water surface of the reservoir. This is an alternate method applicable to western conditions and should not be used on the Atlantic coast where the average rainfall is more than the evaporation from water surface. Reference is made to the literature noted at the end of this article for further discussion of evaporation and the methods of allowing for it. Coefficient of variation The coefficient of variation in annual flows is a coefficient calculated from the records to indicate the relative degree of variability. This term is generally used in statistics and is the square root of the mean of the squares of the percentage variations up and down, without regard to sign, of all the years in the record. *Amer. Soc. C. E., vol. 77, p. 1654. Figure 1 shows a storage diagram on which the storages required for the 95 per cent dry year for various coefficients of variation are plotted. This plotting is made on a logarithmic arithmetic paper which is found most convenient for such diagrams. The curves for C. V. running from 0.20 to 0.45 were derived from records of streams north of the Potomac and east of the Alleghanies and they apply, with proper allowances for the ground storage and evaporation, to all the good data from this region. It is thought that they may be also used on the south Atlantic coast and on the western slope of the Alleghanies, but greater variations are to be expected, and efforts should be made to get local data to check them up or serve as a basis for new curves. No corresponding data have been secured for streams in the Mississippi Valley and in the north central states. The curves now shown may be used tentatively for such cases until local data are obtained, but greater variations must be expected. The curves for a C. V. above 0.50 apply to streams west of the Mississippi, except Oregon and Washington. To this latter group, 90 days storage has been added to cover the variations in rainfall throughout the year. Figures obtained from these curves must be corrected for evaporation and for ground water storage where it exists. Method of expressing storage A storage ratio of 1.00 means that the average storage is equal to the mean flow for one year. A storage ratio of 0.50 means that the storage is equal to half of one year's mean flow; and others in proportion. Another method of stating storage, convenient for certain uses, is in days flow, that is to say as reservoir capacity divided by the daily output that is counted on from the system. Three hundred and sixty-five days storage means one year's actual supply held when reservoirs are full, and if, for example, 70 per cent of the mean flow is available the corresponding storage ratio will be 0.70. Definition of a dry year Estimates are not to be made for the very dryest year. Indeed there is no such thing as the dryest year. Dry years recur at intervals. The dryer the year the longer is the probable interval of |