At intervals of several weeks it will be found necessary to break up the surface of each bed by raking or else to remove a thin coating of clogging material. This should be done after the bed has been rested and dried out, when the surface matting may be taken off without removing much sand. To provide for operating the beds in winter, in the late fall, before the ground has frozen, ridges and furrows should be formed on the surface of the beds, similar to those shown in Fig. 51. The furrows should be two or three feet apart and eight to twelve inches deep. Then when effluent is discharged onto the beds in freezing weather, as it fills the furrows, an ice roof will gradually form, spanning the furrows and protecting the sides and bottoms of the furrows from freezing, especially if a snowfall occurs before severe weather sets in. It will sometimes be found necessary, especially with small beds that are well underdrained, to provide board coverings for the furrows to take the place of the natural ice roofs. [Illustration: FIG. 45.—Five-way Diverting Manhole.] The effluent from the tank should be discharged in such quantities as to flood the entire bed to a depth of from one to two inches, except that some of this effluent will immediately begin to seep into the bed. Respecting the quality and relative fineness of sand suitable for sewage filters, it should be noted that certain empirical methods of measurement have been developed for use in comparing the size and uniformity of particles of various sands. These measures are (1) the “effective size,” and (2) the “uniformity coefficient.” The “effective size” is the size of sand particle expressed in millimetres compared to which ten per cent by weight of the particles in the sample is finer. The “uniformity coefficient” is the ratio of the size of grain which has sixty per cent of the sample finer than itself to the size which has ten per cent finer than itself. Concerning the grades of sand through which sewage may be successfully and properly treated by intermittent filtration, it has been found that the “effective size” should not be less than .20 of a millimetre, nor greater than .50 of a millimetre, and the “uniformity coefficient” should generally be from 1.5 to 3.0, when sewage is applied at the usual rate. If, however, the sand is clean and sharp, but has an “effective size” somewhat smaller than the limit above stated, it may sometimes be found suitable.