stances in the water. For example,when turbidity is largely made up ofcolloidal material and very small parti-cles below O.Sm, as is the case in someclay-bearing water catchments, re-moval efficiencies may range from 0to 40 per cent for raw water turbiditiesbelow 10NTU. Enhanced biologicalactivity within the sandbed definitelycontributes to the improved removalof turbidity.' As the surface waters ofdeveloping countries are becomingincreasingly polluted, improved treat-ment technologies and more stringentstandards should be established. Un-fortunately this is unlikely to happenin the near future, because the govern-ments have little money and there is alack of appropriate monitoring sys-tems. Nevertheless, it is important toaim for low turbidity levels, even ifthis is not specifically indicated innational standards.

Even if high removal efficienciescan be obtained, slow sand filtrationalone cannot always produce drinking-water of a high standard. Raw watersources in many locations in industrial-ized countries are so deteriorated thata combination of treatment processesis required to meet these standards.This is also the case in developingcountries, although these standards arenormally more relaxed.

Reduced efficiencyNormal efficiencies are sometimeshampered by other circumstances.These may interfere with the filtrationand purification processes, leaving toolittle time available between consecu-tive scrapings of the filter to allowadequate maturation. Important inhibit-ing conditions are low temperatures,high turbidities and algal blooms.

Low temperature The efficiency ofslow sand filtration may be reducedwhen the unit is operating under lowtemperatures. E. coli removal will bereduced from 99 per cent at 20°C to50 per cent at 2°C.2

Turbidity Raw water turbidityshould not be above 10NTU for

Gerardo Galvis is Head of CINARA, Universityof Valle, Cali, Colombia, Jan teun Visscher isCo-ordinator for the Slow Sand Filtrationdemonstration project, IRC, The Hague, andBarry Lloyd is Head of the Environmental HealthUnit at the Roben's Institute, University ofSurrey, Guildford, Surrey, GU2 5XH, UK.

duce or inhibit the purification proc-esses in the slow sand filters.

Typical efficienciesTable I presents the typical treatmentefficiencies that slow sand filters canachieve. The reported efficiencies havenormally been achieved in filter unitsoperated at normal hydraulic loadingrates ranging from 0.04 to 0.2m/h attemperatures above 5°C, and in sand-bed depths of greater than O.Sm.

The efficiency rates in Table Icannot always be achieved though,because much depends on the sub-

Multi-stage surface water treatment forcommunity water supply in ColombiaGerardo Galvis, Jan teun Visscher and Barry LloydUsing a multi-barrier treatment for drinking-watergreatly increases the potential of slow sandfilters. Gerardo Galvis, Jan teun Visscher, andBarry Lloyd describe how this treatment processcontributes to a remarkable overall improvementin water quality.

SLOW SAND FIL TRA nON is re-gaining its reputation, both in industri-alized and developing countries, as areliable and highly efficient treatmenttechnology. The filtration process isable to remove most turbidity, andvirtually all harmful entero- (affectingthe intestine) bacteria and entero-viruses and all protozoan cysts. Slowsand filtration is not, however, apanacea for all water quality problemsin all circumstances. Slow sand filtersare inadequate when the levels ofharmful substances in the raw waterare extremely high, and when the rawwater contains substances which re-

The efficiency rates of the treatment systems will depend on the substances inthe water: this river in Peru is overhung with latrines.

'0>-o::;

ai

26 WATERLINES VOL.10 NO.3 JANUARY 1992

Figure 1. Faecal coliform counts per lOOml i1;l three rivers of the AndeanCauca River Valley (sampling period October 1990-May 199/).

Rio Cauca - Pance Colombo -- La Elvira

Figure 2. Turbidity levels (NTU) in three rivers of the Andean CaucaValley. Pance and La Elvira are on the hilly sides of the valley (samplingperiod October 1990-May 1991).

.. "', .. ,,\ "

\ •.,' ''''',; It, '""(

APR MAY

MAR APR MAY

Pance Colombo -- La Elvira

JAN FEB MAR

I

I' \'I' "~ : '. jIL "

~ i" . : ~ :~ II ,'\ ••,,' \ / ••• ;•••' \ ,: ~

I"~,' \/', ," •.•. ,.' ..•.•.,: """: ~,' ~•.' ~, '\ ,At" ',i ~\ II ~:. , \! ....t ~:

~ \:

RioCauca -

OCT NOV DEC JAN FEB

OCT NOV DEC

104

10

5

0.5

Faecalcoliforms 103(CFU/100ml)

Turbidity(NTU)

105

Iron, ManganeseHeavy metals

Cercariae of schistosomaProtozoan cystsTurbidityColourOrganic matter

Table 1: Typical treatment efficiencies of slow sand filters1,2,3

Parameter Typical reduction

Entero-bacteria 90-99% or higher. Coliform removal efficencyis however, reduced by low temperatures,in~reased hydraulic rates, the use of coarse filtersand, shallow sandbeds, decreased contami-nant concentration, and the periodic removal ofthe biological filter skin.Virtually complete removal.99-99.9% removal, even after filter scraping.Generally reduced to less than 1NTU.30-90%, 30%; being the average.COD 30-70%; TOC 15-30%; 50 to 90% oforganic matter such as humic acids, detergents,phenols, and some pesticides and herbicides areremoved.Largely removed.30-90% or even higher.

prolonged periods of time, althoughoccasionally higher peaks of up to50NTU can be accommodated by thefilter units for one or two days withoutmajor increases in head loss. Eventhese short peaks may bury the largenumber of bacterial predators presentin the sand bed and thus reduce theircapacity to remove harmful bacteria.4

Algal blooms Algae may grow in therivers, lakes, or storage reservoirs fromwhich water is brought to slow sandfilters. Most of the algae are retainedby the slow sand filters, but undercertain conditions occasional algalblooms may develop. These massivegrowths may create severe problemssuch as the premature blocking of thefilters, the production of taste andodour in the water, and an increase inthe concentration of soluble and biode-gradable organics in the water. ThepH of the water may rise considerably,causing magnesium hydroxide andcalcium hydroxide to be precipitatedonto the sand grains, thereby eitherblocking the filter or increasing theeffective size of the sand and reducingthe efficiency of the filtration process.

Controlling the algae is difficult, buttwo possible methods are reducing thenutrient contents in the raw water, orcreating a storage system of deepreservoirs in which algae can becontrolled by the exclusion of light.

Colombian projectsIn 1984 the raw water quality of thestreams in the region Valle de Caucain Colombia were studied as part of aresearch and development project ofslow sand filtration.

Two main types of rivers can bedistinguished, highland rivers and low-land rivers. The highland rivers receivewater from relatively small catchmentareas, many of which are facingerosion problems. As a result the waterquality shows short but high peaks ofturbidity when it rains. The lowlandrivers receive both water from thehighland rivers and untreated sewagefrom small and large settlements, andthus have a higher level of contamina-tion but a slower response to rainfall.It was clear that slow sand filters alonecould not cope with the prevailingturbidity levels and faecal coliformcounts in both highland and lowlandrivers (see Figures 1 and 2).

Pilot experiments were established,each with different pre-treatment tech-niques to improve the water qualitybefore it was passed on to the slowsand filters. The results were verypromising, and the pre-treatment sys-tems did remove a considerable partof the suspended solids, reduced faecal

WATERLINES VOL. 10 NO.3 JANUARY 1992 27

The upflow (gravel) filter at the El Retiro treatment plant in Cali.

colifonn counts, and even true colourlevels.s

Selecting and protecting the bestavailable source is far more economi-cal and effective than allowing thewatershed area to deteriorate and thenhaving to rely on advanced treatment.6

Watershed management in developingcountries is as yet not covered byadequate legislation, and there are notenough trained staff to oversee it. It isessential that more collaboration be-tween communities and waste agenciesis encouraged. Other 'social' controlsmay also be necessary, for exampleregulating tree cutting and the use ofpesticides in remove areas, which mayalso mean compensating fanners in thewatershed for lost income, especiallyin regions which have just embarkedon a cash economy.

The subsequent construction of full-scale plants confinned the perfonnanceof the pre-treatment systems and the

slow sand filters. The systems thusconstructed follow the multi-barrierconcept, in which reliance is placedon more than one stage of treatmentto produce safe water for the consum-ers. Together these stages progres-sively remove the contaminants fromraw water and consistently produce asafe and wholesome drinking-water.Ideally, water low in sanitary riskshould then be disinfected with chlo-rine or an alternative, nonnally the lastline of defense and considered a safetybarrier.4 For disinfection to be aneffective safety barrier, the precedingbarriers must remove virtually allhannful micro-organisms and possibleinterfering substances, so that tenninallow-dose disinfection will be an effi-cient safeguard.

Table 2 presents infonnation onseven water treatment systems in theregion Valle del Cauca in Colombia.

PerformanceEach system comprises four differenttreatment steps:1. A conditioning stage, sometimesconsisting of plain sedimentation, butin the newer systems more likely adynamic filter unit. This unit containsa thin layer of fine gravel on top of ashallow bed of coarse gravel. Thewater entering the unit passes over thegravel bed and part of it is drainedthrough the bed to the next treatmentunit, while the rest is returned to theriver. The system will clog when peakturbidities are being received, the flowthrough the bed will then be reduced,and thus low quantities of suspendedsolids will reach the next treatmentunit in the system. In most of thehighland rivers in the Valle del Caucaregion these peaks are of short dura-tion, and cleaning and recommencingoperation of the dynamic filters is onlya matter of a few minutes. Thesesystems are still under development,which explains the different flow rateswhich are being applied.2. Roughing filters, which are mostlyof the single- or multi-stage upflowtype, although two horizontal flowroughing filters are also included.These units are operating at flowvelocities ranging from 0.6 to 1.0m/h.3. Inlet-controlled slow sand filterplants.4. Drop-feed chlorination, which isapplied as a final safety barrier.

The perfonnance of the differentsystems for four important parameters:faecal colifonn, suspended solids, tur-bidity, and true colour were measured.Four of these plants, Colombo, CanasGordas, Retiro, and Javariana, are inan area with average temperatures of24 +- 1DoC, while the other three arein slightly lower temperature zones of

Table 2: Operational parameters of seven water treatment facilities

Pretreatments Slow sand filters

Treatment Flow (lis) Type Length (m) Size range Velocity Filter Filtration(mean) (mm) (mlh) units rate

Ceylan 9.4 URFS 2.0 3-25 0.7 2 0.14La Marina 6.8 URFS 1.8 6-20 0.6 2 0.16Canas Gordas 7.6 OF 0.6 7-25 9.0 3 0.16

URFS 2.0 3-25 0.6EI Retire 8.8 URF 1.0 3-25 0.7 2 0.15Colombo 0.7 OF 0.6 13-25 1.1 2 0.11

URF 1.2 4-25 0.6Cider 0.7 HRF 7.0 5-16 0.8 2 0.15Javeriana 1.5 OF 0.6 5-25 0.8 2 0.08

HRF 4.0 3-16 1.0

URFS =OFURFHRF

28

Upflow roughing filter in seriesDynamic filterUpflow roughing filterHorizontal flow roughing filter

WATERLINES VOL. 10 NO.3 JANUARY 1992

Figure 3. The different methods of treatment adapted to the levels of turbidityand faecal colifarms in the water.

Faecal ColiformsFaecal coliform (CFU/100ml)

100000

Javeriana

Javeriana

Retiro

Retiro

EE Slow sand filtration

as well as industrialized countries, asit will provide an alternative to chemi-cal pre-treatment processes.

ReferencesI. Bellamy, W.O., Si]vennan, G.P., and Hen-

dricks, D.W., Filtration of giardia cysts andother substances, Volume 2 - Slow sandfiltration, US EPA, Cincinnatti, Ohio, 1985.

2. Huisman, L., and Wood, W.E., Slow sandfiltration, WHO, Geneva, ] 974.

3. Hrubec, J. et al. Gedrag van enkele gesubsti-tueerde benzenen, bestrijdingsmiddelen enComplexvonners tijdens langzame zandfil-tratie, H20 VoL24 No.13, pp.348-51.

4. Lloyd, B., The functional microbial ecologyof slow sand filters, PhD thesis, Universityof Surrey, UK, 1974.

5. Galvis, G. and Visscher, J.T., FiltraciollLenta en Arena y Pretratamiento - Tech-n%gia para potabi/izacion, Universidad delValle, Colombia, and IRC, The Hague, 1987and Galvis, G. et ai, Proyecto Integrado deInvestigacion y demonstraction en FiltracionLenta en Arena, CINARA, Cali, ]989.

6. Okun, D.A., 'Best available source', AWWAJournal, March 1991.

Turbidity

f'.a Pre-treatment step

Restrepo Canas Gordas Colombo

Restrepo Canas Gordas ColomboLa Marina

La Marina

Ceylan

.........................................

Ceylan

• Raw river water r2 Pre-treatment step 83 Slow sand filtration

• Raw river water

5

o

10

Turbidity (mean) [NTU]25

10

0.1

100

1000

10000

only requires a very low chlorine doseas an ultimate safety barrier.

Through improved water sourceprotection and pre-treatment systemswhich are now being developed, slowsand filtration can be applied in a muchwider number of settings than earlierthought possible. These findings arevery relevant for developing countries

Filter run length (days)

Demonstraction Head losses (m) Mean Min. Max.projects

Ceylan 0.66 206 206 206La Marina 0.71 34 15 57Canas Gordas 0.65 84 61 115EI Retiro 0.85 85 64 107Colombo 0.28 82 45 105Cider 0.60 178 177 187Javeriana 0.60 203 141 265

18 +- 10°C. Even if the last barrierof low-dose disinfection fails, thewater produced in the Colombiansystems will still represent a lowsanitary risk for the consumers of thetreated water, with average faecalcoliform counts below 1 per 100mlafter slow sand filtration.

Figure 3 shows how the differenttreatment systems adapted to the typeof raw water and the concentration ofimpurities in it. The systems showhigher removal efficiencies for morecontaminated water. In Ceylan theaverage removal of faecal coliformbacteria is 2.6 log units, and inColombo, which receives more con-taminated water, this is 4.4 log units.This aspect is being studied further inthe ongoing research in Colombia,where pilot schemes are now reachingconsistent reduction levels of 5.5 logunits. The pre-treatment systems areas robust as the slow sand filters andare much less vulnerable than thechemical pre-treatment systems. More-over, the treatment efficiency obtainedin the pre-treatment units and the slowsand filters ensures that only a very lowchlorine dose is required for finaldisinfection.

Table 3 presents some data on filterrun periods of the slow sand filterswhich are ranging on average from 2.5to 7 months, with filter rates of someO.15m/h. The only exception is theplant in La Marina, which is havingfilter runs of only one month. Thisplant was one of the first schemesconstructed, and did not benefit fromthe design improvements establishedin later schemes. More important is theamount of algae in the plant, whichmay be the main reason for the shorterruns.

The slow sand filter technology hasworked very well as part of a multi-barrier water treatment system. Underthe conditions in Valle del Cauca, thecombination of slow sand filtration androughing filtration is proving to bevery effective. The levels of turbidity,faecal coliform, and particularly of truecolour were much better than wasexpected. The filtrate from the plants

Table 3: Filtration rate and run lengths of the full scale demonstrationprojects (October 1990-September 1991)

WATERLINES VOL.lO NO.3 JANUARY 1992 29