The method of replacing the washed sand hydraulically seems to have worked better than could have been reasonably anticipated, and the writer believes that this was due, in part, to the excellent method of manipulation described in the paper. It is his feeling, however, that part of the success is attributable to the very low uniformity coefficient of the sand. In other words, the sand grains are nearly all of the same size, due to the character of the stock from which the filter sand was prepared; and, therefore, there is much less opportunity for separation of the sand according to grain sizes than there would be with the filter sand which has been available in most other cases. Filter sand with a uniformity coefficient as low as that obtained at Washington has been rarely available for the construction of sand filters, and while the method of hydraulic return should certainly be considered, it will not be safe to assume that equally favorable results may be obtained with it with sands of high uniformity coefficients until actual favorable experience is obtained.
The writer believes that in calculating the cost of the water used in the plant itself the price chosen by the author, covering only the actual operating expenses of pumping and filtering, is too low. The capacity of the whole Washington Aqueduct system is reduced by whatever quantity is used in this way, and, in calculating the cost of sand handling, the value of the water used should be calculated on a basis which will cover the whole cost of the water, including all capital charges, depreciation, operating expenses, and all costs of every description. On this basis the water used in the sand‑handling operations would probably be worth five or more times the sum mentioned by the author.
The cost of operation of the plant has come within the estimates made in advance, and has certainly been most reasonable. The cost of filter operations has averaged only about 50 cents per million gallons, and is so low that it is obvious that the savings which may be made by introducing further labor‑saving appliances would be relatively small. It will be remembered that ten or fifteen years ago the cost of operating such filters under American conditions was commonly from $2 to $5 per million gallons.
The experiments represented by Tables [17] to [19], inclusive, serve to show that preliminary filtration, or multiple filtration, or any system of mechanical separation is incapable of entirely removing the finer clay particles which cause the residual turbidity in the effluent. They also show that this turbidity may be easily and certainly removed by the application of coagulant to the raw water during the occasional periods when its character is such as to require it.
These general propositions were understood by those responsible for the original design of the plant, as is shown by the author's quotations. These experiments, however, were necessary in order to demonstrate and bring home the conditions to those who thought differently, and who believed that full purification could be obtained by filtration alone, or by double filtration, without recourse to the occasional use of coagulant.
The experiments briefly summarized in [Table 20] are of the greatest interest and importance. Six small filters, otherwise alike and like the large filters, all received the same raw water and were operated at different rates to determine the effect of rate on efficiency.
That the experimental results from the filter operating at the same rate as the large filters were on the whole somewhat inferior to those from the large filters for approximately the same period, may be attributed to the fact that the experimental filter was new while the large filters had been in service for some time and had thereby gained in efficiency. The greatest difference was in the coli results in [Table 20], where it is shown that 24% of the 10‑cu. cm. effluent samples from the experimental filter contained coli, in comparison with only from 1 to 3% of such samples from the main filters.
The results from the experimental filter operating at a rate of 1,000,000 gal. per acre daily may fairly be excluded, as the effluent probably contained more under‑drain bacteria in proportion than filters operated at higher rates. The number of bacteria in the filter operating at a 3,000,000‑gal. rate were 1.7% of those in the applied water; for the filter operating twice as fast, the percentage was 2.4; and, for the one operating more than ten times as fast, was only 3.0; thus indicating a surprisingly small increase in the number of bacteria with increase in rate.
Further and more detailed study by the writer of the unpublished individual results, briefly summarized in [Table 20], confirms the substantial accuracy of the comparison based on the average figures as stated in that table.
It must be kept in mind, in considering these results, that the number of bacteria in each case is made up of two parts, namely, those coming through the filter—which number is presumably greater as the rate is greater—and, second, those coming from harmless growths in the under‑drains and lower parts of the filter—the numbers of which per cubic centimeter are presumably less as the rate is greater—and these two parts, varying in opposite directions, may balance each other, as they seem to do in this case, through a considerable range. It may thus be that the number of bacteria really passing the filter varies much more with the rate than is indicated by the gross results.