materials for con struction of filter box and underdrains, and raw water from a damimpoundment. The ma in act ivities, or operations involved were; procurement, washing and grading of sand and gravel, construction and cha rging of filt er box, preparation of influ ent water, and filtration and data co llection. Based on experience in the local construction industry three main sources of sand were identified; namely, sa nd from rivers (river sa nd), sand which is deposited by storm water (erosion sand), and sand which is found from deposits near seaso nal swa mps (common sand) . A tipper-lorry-load of each type of sand was procured. A tipper load of gravelly soil (locally called 'bush gravel') was also procured. Representative samples of the sand y and gravelly soils were washed t horoughly a nd subj ected to routine tests of speci fic gravity and solubility in 10 per cent hydrochloric acid. These were essentially to confirm what apparently were good quality silica sand s and gravel. . Approximately 1. 1 kg of each soil (air dri ed) was over-dried for 24 hours at I 10° C. One kg of this was weighed into a small handwoven basket (made of palm branches) which had been lined with layers of wire-nettings . This arrangement enabled the so il sample to be washed clean (usin g ta p water) and thus be rid of its soluble components as well as silt a nd sand passing 80 mesh (0.18 mm) . The clean samples were then dried and graded using sieves ranging from 50 mesh (0 .297 mm) through 18 mesh (1.0 mm) for sands and 2.56 mm through 12 .5 mm for gravel. The percentage of each grade found in each was calculated. The standard sieve analyses were performed and D, o, Uc obtained. Filter box was constructed from galvanized iron 2/32" (1 .59 mm) thick. Detai ls of the box are shown in Fig. I . Important features include a float -valve to ensure constant head filtration and an underdrain unit consisting of a galvanized iron pipe perforated on both sides along its length (10 mm holes 10 cm on centres). In placing the filter -media into the box it was decided to keep the height of gravel constant and to vary the height and gradation of sand. Accordingly, in all tests, the bottom 7 .5 cm of tank height were filled with gravel of sizes passing 12.5 mm but retained on 5.6 mm sieves, and the next 15.0 cm with gravel of sizes passing 5.6 mm but retained on 2.56 mm sieves . Depths of sand used were varied between 60 cm and 75 cm; and sand gradations included varied combinations of particle sizes sorted from amongst those passsing 1.0 mm and retained on 0.30 mm sieves. One important constraint in the use of the filter box as constructed was that it had no facilities for backwashing. To ensure a reasonable level of particle segregation (smallest particles at top and coarsest at bottom) every batch of sand taken from the 0.3 mm-1 .0 mm stock was first regraded by sieve nos 18, 20, 30, 40 and 50 (i.e. 1.0, 0.85, 0.59, 0.425 and 0.30 mm) into near uni-sized subbatches . The filter box was half-filled with water and rocked gently as the media were . added in small amounts (starting with the coarsest sub-batch). TABLE 1. CHARACTERISTICS OF FILTER MEDIA FROM DIFFERENT SOURCES Colour Odour River sand Erosion Sand ·common' Moist Dry Light brown Grey Brown Brown Light brown Grey wit h brown specs Moist None No ne dry No ne None of decomposing vegetables None 2.59-2 .66 2.63-2.65 2.61 -2.67 2.65-2.69 Speci fi c gravi ty ranges (3 tests) Bush gravel Sand Dark grey Dark grey None None mean 2.66 2.64 2.65 2.66 OJo yield : 0.18 < d < 0.30 mm 0.30 19 12.4 10.6 12 .0 13.4 9. 1 8. 1 6.0 0.0 0.0 13 .8 14.0 10.5 17.0 14.2 9.3 8.9 0.9 0.3 11.7 9.0 4.2 9.4 12.3 16.4 2.5 0. 2 0.0 0.1 0.2 2.4 4.0 6.0 19.3 41.2 4.4 2.0 Useable sand (.3-1.0) Useable gravel (4 .75- 19.0) 45 . I OJo 6.00/o 55.70/o 10. I OJo 34.90/o 2.70/o 12.60/o 45 .60/o T otal useable mate ria ls (0.3-19.0)• 59.20/o 75 .I OJo 54.00Jo 79.60/o "' • The pe rce ntage shown for 'Total useable materials' include those in the range I.0 < d < 4.75 whi ch is probably on ly marginally useable. The influent water for the filter was prepared as follows: A half-ta nkerfull of raw water from the Opa Water Works (the Water Works of the University of Ife) was drawn into a tank just outside the laboratory. Preliminary laboratory experiments had indicated that the optimum alum dosage for coagulating the water was about 50 mg/ L if the pH was first adjusted to about 8.5 (50 mg/ L of alum lowered the pH from 8.5 to approximately 7 .0). Rapid mixing had also been found to be good in the velocity gradient (G) range 200 sec- • to 600 sec-•, provided that for the high velocity gradients a good amount of flocculation (slow mixing, G of about 25 sec- ') was mandatory. Measurement of velocity gradients was by use of a Master Servodyne (Product of Cole-Parmer Instrument and Equipment Company, Chicago, Illinois, U .S.A.), using Calibration curves similar to those used by O gedengbe (1975). On the basis of the preliminary tests, coupled with the desire to have a fairly turbid filter influent, the influent water was prepared by using a n alum dosage of 50 mg/ L ; rapid mixing at a G-value of 500 sec-• for 2 minutes; providing no slow mixing; and settling for 10 minutes. The supernatant was collected into a box from which a one Horsepower pump lifted the water into the filter box through the float-valve . The effluent control valve on the filter box was set to pass between 100 litres/ mini m' and 150 litres/ mini m', with filter media in place (at the beginning of the filter run) . This choice was based on the fact that rapid filter s are operated at flow rates ranging from 2 gpm/ft' (81 .6 litres/ mini m') and 6 gpm/ft' (244 .8 litres/ mini m' ) as many published works have shown (Cleasby and Baumann, 1962; Fair et al, 1971, ASCE et al, 1969; and so on). During filtration , the turbidity of the influent water was measured every 30 minutes (or less when deemed necessary) on a HACH Turbidimeter, m·o del 2100A. The turbidity of the efflu ent was similarly measured . Flow rates were measured at about the same time that turbidity samples were taken . Headloss readings were also taken as the difference between the water levels at the top and bottom headloss ports. Whenever each run was terminated , the tank was washed and the media were replaced by the next batch of clean, graded gravel and sand. A filter run was terminated (generally) when the turbidity increased to about 1.5 NTU or when head loss reached about 100 cm. RESULTS AND DISCUSSIONS Table I shows the physical characteristics of sand and gravel from various sources and the ranges of sizes c,f their particles after those smaller than 50 mesh (0.30 cm) had been washed off. It can be seen from the table that, of the three types, erosion sandstock provided the highest sand yield (55 per cent useable sand, 10 per cent useable gravel). The river sandstock yielded about 45 per cent sand , 6 per cent gravel; and the 'common' sand 35 per cent sand and about 3 per ce nt gravel. On the other hand, the 'bush gravel' yielded about 13 per cent useable sand and 48 per cent useable gravel. Furthermore, each of the four stocks yielded between 8 per cent and 20 per cent of media in the size range 1.0 mm to 4.75 mm which, though not strictly filter sand nor filter gravel, could be useful nonetheless . When particles of sizes equal to or larger than 1.0 mm were eliminated from the stock sa nd (as well as sizes smaller than 0.30 mm), the analysis of the resulting stocks were as shown in Table 2. The values of the uniformity coefficient (Uc) for the river sand, the erosion sand and the common sand were 2.67, 2.69 and 2.95 respectively. The results obtained when a 60 cm depth of sand representative of each of those described in the preceding section (passing 1.0 mm sieves but retained on 0.30 mm sieves) was WATER December, 1982 21