As far as the question of capacity is concerned, these diagrams indicate that a filter in normal condition may safely be raked once. It is believed that the constantly decreasing capacity shown in Group D is not due so much to the rakings as to the small quantities of sand removed at the alternate scrapings, and therefore it would not be proper to condemn this method of treatment without a further trial in which this defect was remedied. This view seems to be supported by the results of Group B. The low but approximately constant capacity there shown would undoubtedly have been higher if a greater depth of sand had been removed each time.



Quality of the Effluent.—The averages given in Table 29 show but little difference in the bacterial contents of the effluents from the four groups of filters. All are entirely satisfactory, and the differences in favor of one method or another are small. In looking for possible differences in the quality of the effluents from the four groups, it was thought that such differences might be most apparent at a time when the entire plant was working under the most adverse conditions. The bacterial counts, therefore, were summarized for the period from December 23d, 1907, to January 6th, 1908, inclusive, following a period of high turbidity and high bacteria in the raw water, with results as follows:

Group....... A B C D Maximum..... 204 178 189 206 Minimum..... 61 45 62 57 Average..... 120 107 104 155

The following is a summary of the turbidity results for a similar period:

Group....... A B C D Maximum..... 10.8 11.7 8.7 9.3 Minimum..... 6.7 4.7 6.2 5.7 Average..... 8.7 8.3 7.2 7.9

These numbers, though high, do not show any significant differences. All the averages for each group are less than the lowest maximum, and all are greater than the highest minimum, and therefore vary less than do the individual filters, from other causes, within the different groups.

Future Capacity of the Filters.—An indication of the dangers which might affect the future capacity of the filters was shown in the above discussion of the present capacity. A more effective way of showing this was obtained by a study of the initial resistances or losses of head in the four groups. A filter kept in ideal condition would show no increase in this initial loss of head from one run to the next. If there is such an increase, it means that at some future time measures more heroic than ordinarily used would be necessary to restore the proper capacity.

The average initial losses of head for the different groups are plotted on the diagram, Figure 13. Group A shows an initial loss of head, increasing gradually but slightly during more than two years of service. In Group B the initial loss of head increased in a manner similar to that in Group A, up to the time of the beginning of these experiments; after which the increase becomes more rapid. Groups C and D show conditions generally similar to Group B, with some variations which are self-explanatory.

Conclusions.—The quality of the effluents from all four groups was satisfactory, and no consistent difference was apparent in favor of one or another method of treatment. The method pursued with Group C was entirely insufficient to maintain the capacity indefinitely. The methods pursued in Groups B and D were both insufficient, but would have been more effective if a greater depth of sand had been removed. The costs of treatment of Groups B and D were less than for Group A. It appears, then, that a treatment which would be more economical than the old method of Group A, and would still maintain the proper capacity, would be one similar to that of Groups B or D, with the removal of a quantity of sand greater than was done in the case of these two groups, but less than in the old method.



At the time the above results were summarized, it was proposed to proceed with the filter treatment along the lines just mentioned. The writer did not have an opportunity to study the subsequent results, as he was transferred to other work. A statement by the author of any new facts that may have come to light in this connection would be of interest.

Mention should be made, too, of another expedient that was used to hasten the restoration of the capacity of a filter, which proved to be a most useful one. The removal of the scraped sand from a filter was a matter of a good many hours' work, under the most favorable conditions. To get the filters quickly into service again, the dirty sand in a number of them was simply scraped from the surface, heaped into piles, and left there; then the water was turned in, and the filter was started again. This was done with some hesitation at first for fear the presence of the piles of dirty sand might cause high bacterial counts in the effluents of those filters. No such effect was observed, however, the counts being entirely normal throughout. The writer subsequently found the same treatment being applied as an emergency measure at the Torresdale plant, in Philadelphia, and, through the courtesy of the Chief Engineer of the Bureau of Filtration, was furnished with the bacterial counts through a number of runs made under these conditions, and there, too, the results were entirely normal.

There was practically no economy in this method, as the sand had ultimately to be ejected and washed. The piling up of the sand had the effect of reducing the effective filtering area by a small percentage, with a corresponding increase in the actual rate of filtration, but this was of trifling importance. The great benefit derived from the method was the saving of time in getting a filter back into service after scraping, and in this respect it was very valuable.

Physical Theory of Purification of Water by Slow Sand Filters.

The first and most natural conception of the action of a sand filter is that the removal of impurities is effected by a straining action. This, of course, is perfectly true as far as it relates to a large part of the visible impurities. Much of this is gross enough to be intercepted and held at the surface of the sand. This very straining action is an accumulative one. After a quantity of suspended matter thus strained out mats itself on the surface of the sand, it in turn becomes a strainer, even better adapted than the clean sand surface which supports it for the removal of suspended matter from the water.

This, however, cannot explain certain features of the purification of water by a layer of sand. The removal of color, the reduction of nitrates, and certain other changes in the organic content of the water have for a long time been recognized as due to a bio-chemical action carried on by certain bacteria in the sand. Both the straining action and this bio-chemical action are not all-sufficient for the explanation of certain phenomena, and it has been recognized, too, that sedimentation in the pores of the sand played a large part in the purification process in those cases in which it was apparent that the biological agencies were not the chief ones.

In the purification of water containing only insignificant quantities of suspended matter, but a relatively large amount of unstable organic matter, it will be conceded at once that the chief factor in the purification is the nitrification produced by the bacteria in the upper layers of the sand. On the other hand, the purification by sand filters of a hypothetical water containing no organic matter, but only finely-divided mineral matter in suspension, could take place only by the physical deposition of the particles upon the sand grains. Between these two extremes lie all classes of water. In all problems of water purification by filtration through sand, both these factors—biological action and sedimentation—play their parts, assisting and supplementing each other, the relative importance of one factor or the other depending on the place of the particular water in question on the scale between the two extreme conditions just mentioned.

In Mr. Hazen's paper on "Sedimentation"[1] there is an interesting development of the theory of the removal of suspended matter by sedimentation in the pores of a layer of sand. The factors influencing this removal are the rate of filtration, the effective size of the sand, and the temperature of the water. For the conditions at the Washington plant, it may be assumed that the first two of these factors are constant. The third factor, however, varies through wide limits, and the observations on the turbidity removal, and on the different phases of the filter operation of which the turbidity of the water is a factor under varying temperature conditions, together with the known relations between hydraulic values and temperatures of water, furnished good substantiative evidence that this highly-induced sedimentation may be a considerable factor in the purification of the water as effected at this plant. This temperature relation, briefly stated, is as follows: For particles of a size so small that the viscosity of the water is the controlling factor in determining the velocity of their subsidence in still water, that velocity will vary directly as (T + 10) / 60, in which T is the temperature, in degrees, Fahrenheit. That is, when the temperature of the water is between 70 deg. and 80 deg. Fahr., a particle will settle with twice the velocity it would have if the water were near the freezing point.

[Footnote 1: Transactions, Am. Soc. C. E., Vol. LIII, p. 59.]

The layer of sand in a slow sand filter may be considered as a very great number of small sedimentation basins communicating one with another, not in the manner of basins connected in series, but rather, as Mr. Hazen has expressed it, as a long series of compartments connected at one side only with a passageway in which a current is maintained. In any section of the sand layer there are areas through which the water passes with a velocity much greater than its mean velocity through the total area of voids, while there are other areas in which the velocity is very much less, perhaps in an almost quiescent state from time to time, greatly favoring the deposition of particles, but with a gentle intermittent circulation, displacing the settled or partly-settled water and supplying from the main currents water containing more suspended matter particles to be removed. There is thus a considerable percentage of the total volume of voids in which the water is subjected to very favorable conditions for sedimentation, almost perfect stillness and an exceedingly small distance for a particle to settle before it strikes bottom on the surface of a grain of sand.

If sedimentation were the predominating factor in the purification of the water, we would then expect to find the following phenomena in the operation of the filters: A more rapid deposition of a given amount of sediment under summer temperature conditions than under winter, as the water passes through the sand, and therefore, for the former condition of higher temperature:

(a) A greater concentration of this turbidity-producing material in the top layer of sand, or, in other words, a thinner sand layer to be removed in scraping if all the dirty sand is removed;

(b) Because of the greater concentration, a greater rate of Increase of the loss of head, and consequently shorter periods of service between scrapings;

(c) A higher limit for turbidity in the water applied to the filter to produce a given turbidity in the effluent.

Table 30 Service Periods and Scraping Depths for Runs Ending In Various Months; Covering Entire Period, October 1st, 1905, To March 1st, 1907. ================================================== Average Average Mean Number period of depth of sand temperature, Month. of service removed, in in degrees, filters. in days. inches. Fahrenheit. - - - - January 13 75 2.09 39 February 6 98 2.46 37 March 5 130 2.66 41 April 8 149 2.96 53 May 7 130 2.80 67 June 11 124 2.35 77 July 17 70 2.12 81 August 2 49 1.98 80 September 5 73 2.48 76 October 37 70 1.56 64 November 20 42 0.81 49 December 14 57 0.94 40 ======================================================

The operation of this plant during the first year and a half offered an excellent opportunity for the study of sedimentation in the sand, and the data in Table 30 are presented to show that certain of the phenomena of filter operation observed during this period seem to be fairly explicable by the physical theory of purification. These data are given only for the period of operation before the summer of 1907. At that time the experiments in filter cleaning already described were begun. Before that time, whenever a filter had been cleaned, all the discolored sand had been removed, leaving for the following run a new sand surface substantially in the perfect condition of a newly-constructed filter. After that time the experimental methods of cleaning, and the new routine adopted as a result thereof, interfered with the tracing of the evidence as clearly as during the earlier periods.



Table 30 and the corresponding diagram, Figure 14, show the general variations in the length of runs and depth of penetration, with the seasonal temperature changes. The increase in length of runs and quantity of sand removed under low temperature conditions is very marked. There is, however, a secondary maximum which appears, as the diagram shows, where a minimum for the year would be expected. This may have been an irregularity occurring this one year, which will not appear in the average of several years, and caused by some factor which has escaped observation. A careful analysis of the data at hand fails to show any explanation for it. It may exist in some of the little-understood biological actions which have their maximum effect under warm-water conditions, or it may be due—in some obscure way—to the liberation of air under the surface of the sand, accumulating with pressure enough to break the surface at innumerable points, thereby reducing the loss of head and extending the period of service. Some evidence was observed pointing to this explanation, but it was never conclusively proven.

The general effect of temperature changes on the rapidity of removal of the sediment and its consequent concentration in the sand layer, however, seems plainly evident.

In corroboration of the third point mentioned in the theoretical consideration of turbidity removal in the filters, the daily turbidities of the filtered water have been classified and summarized for different turbidities in the applied water, and also for different temperatures. The average turbidities thus obtained are given in Table 31.

Table 31 Turbidity in Filtered Water at Different Temperatures Produced by Given Turbidity in Applied Water. ========+=============================================== Turbidity of Temperature, in Degrees, Fahrenheit. applied -+ -+ -+ -+ - water. 40 40 - 50 50 - 60 60 - 70 70 + -+ -+ -+ -+ - 20 1.8 1.3 1.2 1.5 1.7 20-40 4.8 5.0 3.5 3.0 2.6 40-60 7.9 6.9 5.4 ... 3.7 60-80 10.7 7.7 ... ... 5.4 80-100 11.3 ... ... ... ... 100 ... ... ... ... 12.0[1] =========+========+========+========+========+========

[Footnote 1: For an average turbidity = 150. approximately.]

The influence of the temperature of the water on the turbidity of the effluent is very pronounced. For a temperature of less than 40 deg. Fahr. (actual average temperature about 35 deg.), the turbidity of the filtered water for a given turbidity of the applied water is practically twice as great as for a temperature greater than 70 deg. (actual average temperature about 75 deg.). This fact fits in very nicely with the influence of temperature on sedimentation. Referring again to this temperature relation, as set forth on a previous page, the hydraulic subsiding value of a particle in water, of a size so small that viscosity is the controlling factor in its downward velocity, is approximately twice as great at 75 deg. as at 35 degrees. We would then expect to find that, in order to obtain a given turbidity in the filtered water, a raw water may be applied at 75 deg., having twice the turbidity of the water applied at 35 deg., to produce the same turbidity; and further, as the turbidity of the filtered water, for a given temperature condition, varies quite directly in proportion to the turbidity in the applied water, it follows that an applied water of given turbidity will produce an effluent at 35 deg. with a turbidity twice as great as at 75 degrees. This is quite in accordance with the facts obtained in actual operation, as indicated on the diagram, Figure 15.

Preliminary Treatment of the Water.—The most striking features of the bacterial results given in Table 4 are, first, the uniformly low numbers of bacteria in the filtered water during perhaps 8 or 9 months of the year, and the increase in numbers each winter. This is shown clearly in the analysis of bacterial counts in Table 32.

Table 32 Classification of Daily Bacterial Counts in the Filtered-Water Reservoir During the Period, November 1st, 1905, to February 1st, 1908. ========================================================== Bacterial count between: No. of days. Percentage of whole. 0 and 20 per cu. cm. 291 41.0 20 and 40 per " " 245 34.6 40 and 60 per " " 63 8.9 60 and 80 per " " 30 4.2 80 and 100 per " " 28 4.0 92.7 100 and 200 per " " 29 4.1 200 and 300 per " " 13 1.8 300 and 500 per " " 5 0.7 500 and 1000 per " " 5 0.7 7.3 100.0 ==========================================================

The tests for Bacillus Coli in Table 5 show results which correspond closely to these, with this organism detected only infrequently, except during the periods of high bacteria, and both of these are parallel to the turbidity variations in the filtered water. These variations follow closely the variations in the turbidity and in the bacterial content of the water applied to the filters.

By all standards of excellence, the sanitary quality of the water during the greater part of the time is beyond criticism. In view of the close parallelism of turbidity and bacterial results in the applied and in the filtered water, it is entirely logical to conclude that, if the quality of the applied water could be maintained continually through the winter as good as, or better than, it is during the summer, then the filtered water would be of the perfect sanitary quality desired throughout the entire year.

This was all foreseen ten years ago, when Messrs. Hering, Fuller, and Hazen recommended auxiliary works for preliminary treatment of the supply, although, as the author states, these works were not provided for in the original construction. As prejudice against the use of a coagulant seemed to be at the bottom of the opposition to the preliminary treatment, a campaign of education bearing on this point was instituted, in addition to the systematic studies of different preliminary methods to which the author refers. As a result of the combined efforts of all those interested in promoting this improvement, an appropriation was finally made for the work in 1910. The coagulating plant has since been built, and the writer is informed that coagulation was tried on a working scale a short time ago during a period of high turbidity. A statement of the results of this treatment on the purification of the water in the reservoir system and in the filter plant would be of great interest.



Hydraulic Replacing of Filter Sand.—The author has adopted a method of replacing clean sand in the filters which will commend itself to engineers as containing possibilities of economy in operation. The first experiments in the development of this method at the Washington plant were carried out some three years ago, while the writer was still there. Substantially the same methods were used then as are described in this paper, but examination of the sand layer by cutting vertically downward through it after re-sanding in this manner showed such a persistent tendency toward the segregation of the coarse material as to hold out rather discouraging promises of success. The greatest degree of separation seemed to be caused by the wash of the stream discharging sand on the surface. It was observed that, near the point where the velocity of the stream was practically destroyed, there seemed to be a tendency to scour away the fine sand and leave the coarse material by itself, and pockets of this kind were found at many points throughout the sand layer. The author states that, in the recent treatment of the filters by this method, there has been no apparent tendency for the materials to separate into different sizes, and it is fortunate if this work can be done in such a manner as to avoid this separation entirely.

It may be questioned whether a certain amount of segregation of the materials will make any practical difference in the efficiency of a filter. In all probability this depends on the degree of the segregation, the quantity of pollution in the water to be filtered, the rate of filtration, and the uniformity of methods followed in the operation, etc. For an applied water as excellent in quality as that of the Washington City Reservoir during favorable summer conditions, a considerable degree of segregation might exist without producing any diminution in efficiency. For a badly polluted water, however, such as the applied water at this plant during certain winter periods, or the water of a great many other polluted supplies, it might be found that even a slight lack of homogeneity in the sand might make an appreciable difference in the results of filtration.

As a result of the experiments herein described, however, this method may be applied at other plants where conditions seem to warrant it, with a largely increased measure of confidence; although, as in the case of the adoption of any new or radical departure, that confidence must not be permitted to foster contempt of the old and tried methods, but its operation must be watched with the utmost caution, until long experience shall have demonstrated its perfect suitability and defined its limitations.

E. D. Hardy, M. Am. Soc. C. E. (by letter).—It was not the writer's original intention to enter into a discussion of either the theory of water purification or of the experimental work on sand handling, but simply to present the main results of operation largely in tabular form. He is gratified, however, to have these sides of the question so ably brought out in Mr. Longley's discussion.

Mr. Hazen referred to the inferior efficiencies of the experimental filters for rate studies (as shown in Table 20) in the removal of the B. Coli from the water tested. This inferiority is really less than the figures in the table would indicate, as the tests for the experimental filters were presumptive only (as shown by the note at the foot of Table 20), while those for the main filters were carried through all the confirmatory steps.

From experiments[1] made by Messrs. Longley and Baton in the writer's office, it would seem reasonable to assume that about one-half of the positive results, would have been eliminated had the confirmatory steps been taken. In other words, the figures showing the number of positive tests for B. Coli in Table 20 should be divided by two when comparing them with corresponding ones for the main filters.

[Footnote 1: Published in the Journal of Infectious Diseases, Vol. 4, No. 3, June, 1907.]

Mr. Knowles seems somewhat apprehensive regarding the methods described in the paper of restoring the capacity of the filters by raking, and replacing sand by the hydraulic method, and yet, from Mr. Johnson's discussion, it would seem that the practice of raking filters between scrapings had recently been adopted at the Pittsburg plant.

Before the practice of raking was finally adopted as a part of the routine filter operation, the subject was given a great deal of thought and study, as may be seen by referring to Mr. Longley's discussion.

The re-sanding has been done by the hydraulic method, for nearly two years, and, as far as the writer is able to judge, this method has been more economical and also more satisfactory in every way than the old one. As Mr. Hazen states, this does not prove that the hydraulic method would be as satisfactory for other filter plants and other grades of sand. The elevated sand bins at the Washington plant fit in well with this scheme, and save the expense of one shoveling of the sand; and the low uniformity coefficient of the sand is favorable in decreasing its tendency to separate into pockets or strata of coarse and fine sand. The method of washing is also well adapted to this method of re-sanding, as the sand is made very clean in its passage through the washers and storage bins. The hydraulic method of replacing sand tends to make it cleaner still, because any clay which may be left in the sand is constantly being carried away over the weir and out of the bed, to the sewer. Sand replaced by the hydraulic method is much more compact than when replaced by other methods, and consequently the depth of penetration of mud in a filter thus re-sanded is less. Careful tests of the effluents from filters which have been re-sanded by the two methods have invariably shown the superiority of the hydraulic method.

The experiment of replacing sand by water, referred to by Mr. Longley, was not considered a success at the time, and the method was abandoned for about a year. At that time an attempt was made to complete the re-sanding of a filter which had been nearly completed by the old method. The precaution of filling the filter with water was not taken, nor was any special device used for distributing the sand. When this method was again taken up, various experiments were tried before the present method was adopted.

Mr. Whipple's remarks on the results from the operation of filters under winter conditions are very interesting, and, considering his standing as an authority in such matters, they are worth careful consideration.

In the operation of the Washington plant, it has always been noticeable that the results were much poorer in winter than in summer. In fact, nearly all the unsatisfactory water which has been delivered to the city mains has been supplied during the winter months. On the other hand, the typhoid death rate has always been comparatively low in cold weather. These facts would seem to indicate that the water supply was not responsible for the typhoid conditions.

THE END