nation showed repeatedly more than 1,000 and 1,500 standard
units of _melosira_ per cu. cm., and one count showed nearly 3,000
standard units.

Several expedients were tried in an effort to restore the rapidly
decreasing capacity of the filters. One of the earlier conjectures
as to the cause of the trouble was that it might be due to the
accumulation of large quantities of air under the surface of the
sand, as air had been observed bubbling up through the sand,
especially in filters which had been in service for some time. The
expedient was tried, therefore, of draining the water out of the
sand and then re-filling the filter in the usual manner from below,
in the hope of driving out the entrained air. Presumably this
treatment got rid of the air, but it did not restore the capacity of
the filter, as the point of maximum resistance was in the surface of
the sand and not below it.

As the author states, raking the filters was tried and found to give
results which were satisfactory enough to meet the emergencies
already referred to. When the filters were first put in operation,
in the fall of 1905, the method of bringing back the capacity of a
filter after the end of a run was to remove all the dirty sand to a
depth determined by the marked discoloration caused by the
penetration of the clay turbidity. This sometimes necessitated the
removal of large quantities of sand at a cleaning, as the turbidity
was exceedingly fine, and penetrated at times to a depth of 3 or 4
in.

With the idea of effecting an economy in the cost of cleaning the
filters, a schedule of experiments was arranged shortly before July
1st, 1907. The general object of the experiments was to determine,
first, the relative costs of all different methods tried; second,
whether the removal of only a thin layer of sand, or the mere
breaking up of the surface of the sand by thorough raking, would
give the filter its proper capacity for the succeeding run; third,
whether the filters under these treatments would maintain a high
standard of quality in the effluents; fourth, whether the continued
application of any less thorough method than the one then in use
might materially affect the future capacity of the filters.

To this end the filters were divided into four groups which, during
a period of about six months, were subjected to treatments as
follows:

Group _A._--Filters scraped deep at the end of each run;
Group _B._--Filters scraped light at