Removal of Bacteriophages MS2and ΦX174 during Transport in aSandy Anoxic AquiferP A U L W . J . J . V A N D E R W I E L E N , *W I E L J . M . K . S E N D E N , A N DG E R T J A N M E D E M A

Kiwa Water Research, P.O. Box 1072, 3433BB Nieuwegein,The Netherlands

Received January 16, 2008. Revised manuscript receivedMarch 27, 2008. Accepted April 1, 2008.

The objectives of our study were to determine (i) removal ofbacteriophage MS2 and ΦX174, as surrogates for humanpathogenic viruses, in an anoxic aquifer and (ii) the safe lengthof the microbial protection zone in anoxic aquifers. 3.5 Logunits of MS2 were removed by adsorption and inactivation during63 days residence time, which was 1.4 log units lower thanremoval of ΦX174. These removal rates were considerably lowerthan previously reported for MS2 and ΦX174 in oxic aquifersand consequently longer protection zones around anoxic aquifersmight be needed. Therefore, the observed log removal ofMS2 was used in a risk assessment approach to determinethe required safe length of the microbial protection zone. In caseof a leaking sewer in the vicinity of a well in an anoxicaquifer, the risk assessment demonstrated that a microbialprotection zone of 110 m may be needed to meet a risk of infectionof 10-4 persons per year. This length can be two to threetimes larger than the length of the protection zone currentlyused in a number of countries.

IntroductionGroundwater abstraction wells are protected against mi-crobial contamination by microbial protection zones aroundthe abstraction well. The length of the microbial protectionzone differs among countries, but mostly a water residencetime of at least 50-60 days between the edge of the zone andabstraction well is used (1, 2). The 50-60 days water residencetime is based on research in the 1930s by Knorr, who showedthat removal of Escherichia coli during 50- 60 days transportin an aquifer was sufficient to protect abstraction wells againstbreakthrough (3). However, even at that time Knorr remarkedthat it was uncertain whether 60 days residence timeprotected abstraction wells against virus contamination.

Since the 1930s, knowledge about microbial transportthrough soil has increased considerably and many studieshave shown that bacteria (including E. coli) were removedat higher rates than viruses during transport throughsaturated soil (reviewed in refs 4, 5). Consequently, fieldstudies were conducted to study removal of bacteriophages,as surrogates for human pathogenic viruses, in sandy aquifers(6–12). These studies all demonstrated that high removalrates (up to 8 log removal) were achieved within 30 daysresidence time in the aquifer. As a result, a microbialprotection zone based on 50-60 days residence time would

be sufficient to protect the abstraction well against virusbreakthrough from a contamination source.

Recently, the length of the protection zone around anumber of groundwater abstraction wells in The Netherlandswas calculated using a modeling approach (13, 14). Theconclusion from these studies was that a microbial protectionzone based on 60 days residence time, the guideline for themicrobial protection zone around groundwater abstractionwells in The Netherlands, protected abstraction wells in oxicaquifers sufficiently. In contrast, the microbial protectionzone around anoxic aquifers should be based on 2-2.5 yearsresidence time to stay with 95% certainty under an infectionrisk of 1 out of 10,000 persons per year; the tolerable infectionrisk for drinking water according to the current Dutchlegislation (15). A sensitivity analysis showed that the virustransport model was most sensitive for inactivation rate andcollision efficiency of viruses (13, 14). The value of bothparameters under anoxic conditions is unknown and wasassumed, because no field studies have investigated virustransport entirely under anoxic conditions. Hence, the modelresults of both studies are more indicative rather than definite.

The objective of the current study was to determineremoval of bacteriophage MS2 and ΦX174 in an anoxic sandyaquifer. To achieve this objective, anoxic groundwatercontaining high titers of MS2 and ΦX174, as surrogates forhuman pathogenic viruses, was injected under anoxicconditions in an anoxic aquifer with a relatively high pH,and removal of both bacteriophages during transport wasdetermined. The results obtained in our field study werealso used to determine the required length of the microbialprotection zone around groundwater abstraction wells undera worst case scenario using a risk assessment approach.

Materials and MethodsField Location. The removal of bacteriophages in an anoxicaquifer was studied at a field location in The Netherlands.The aquifer is positioned at 9 m below ground level andconsists of moderately coarse sand. A 9 m confining layer ofclay and loam overlies the aquifer. An abstraction well wasplaced in the aquifer at a depth of 10 to 15 m below groundlevel (Figure S1), and groundwater was abstracted with 11.32m3 h-1. One injection well and two monitoring wells, withthe screen depth at 11 to 13 m below surface level, wereplaced at a horizontal distance of 20.5, 29.9, and 37.7 m fromthe abstraction well, in line with the groundwater flowdirection. The groundwater is anoxic (O2 < 0.5 mg L-1; NO3

< 0.5 mg L-1; Fe g 0.1 mg L-1, and Mn g 0.1 mg L-1; (16)),has a relatively high pH of 7.5, and a temperature of 12.9 °C.The majority (97%) of the aquifer consists of sand; theremaining 3% is silt (2.4%) and clay (0.6%). The porosity ofthe aquifer is 0.32 and the geometric mean of the grain sizeis 405 µm. The percentage organic matter in the aquifer islow (∼ 0.15%) and the cation exchange capacity variesbetween 5.6 and 19.3 meq kg-1.

Seeding Experiments. Before sodium bromide was in-jected as a conservative tracer, the abstraction well was inoperation for two months and had abstracted a volume of15,000 m3 water. Bromide tracer injectate was prepared bydissolving 500 g NaBr in 1 L of demineralized water. Thetracer was diluted to 809 mg L-1 bromide with anoxicgroundwater in a seeding tank under a continuous flow ofnitrogen gas to keep the solution anoxic. The bromide tracerwas injected in the injection well at 12 m depth with a flowof 18.2 L h-1 for 96 h, resulting in a total injected volume of1747 L of tracer solution. During injection of the bromidetracer, the seeding tank was kept under anoxic conditions as

* Corresponding author phone: +31 (0)306069642; fax: +31-(030)6061165; e-mail: paul.van.der.wielen@kiwa.nl.

Environ. Sci. Technol. 2008, 42, 4589–4594

10.1021/es800156c CCC: $40.75 2008 American Chemical Society VOL. 42, NO. 12, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 4589

Published on Web 05/14/2008

well. The two monitoring wells and abstraction well weresampled for the next 93 days.

High titer solutions of bacteriophage MS2 and ΦX174 wereobtained from GAP Enviromicrobial Services (London,Canada). One liter of MS2 and 1 L of ΦX174 solutions werediluted together in a seeding tank with anoxic groundwaterunder anoxic conditions. After dilution, MS2 concentrationin the seeding tank was 6.9 × 109 pfu mL-1, whereas ΦX174concentration was 2.5 × 106 pfu mL-1. The bacteriophagesolution was injected under anoxic conditions for 96 h at12 m depth in the injection well with a flow of 19.8 L h-1,resulting in a total volume of 1901 L of bacteriophage solution.The injection well, two monitoring wells, and abstractionwell were sampled for the next 163 days.

Analytical Methods. Samples were taken from injectionand monitoring wells using a mobile pump. To samplegroundwater, two times the volume of stagnant water in thewell was pumped out before a sample was taken for bromideor bacteriophage analyses. Samples from the abstraction wellwere taken at a tap located on the abstraction well. Becauselow concentrations of bacteriophages were expected at theabstraction well, volumes up to 10 L were analyzed for MS2and ΦX174.

Bromide was analyzed using ion chromatography. Bac-teriophages were determined by enumerating plaque formingunits (pfu) using the agar overlay technique. BacteriophageMS2 was determined as described in ISO 10705-1 usingSalmonella typhimurium WG49 as host strain, and bacte-riophage ΦX174 was determined according to ISO 10705-2using E. coli WG5 as host strain.

Inactivation Rate of Free Bacteriophages in Ground-water. The inactivation rate of free bacteriophages ingroundwater was determined by sampling (i) water from theseeding tank, (ii) groundwater from the first monitoring wellduring maximum breakthrough of the bacteriophages (18days after bacteriophages were injected in the injection well),and (iii) groundwater from the second monitoring well 43days after maximum breakthrough of both bacteriophages(80 days after bacteriophages were injected in the injectionwell). The three water samples were stored in bottles understrict anoxic conditions at 12 °C, conditions identical to theredox and temperature conditions of the groundwater. For100 days, the bottles with anoxic groundwater were sampledweekly and numbers of bacteriophage MS2 and ΦX174 were

determined using the methods as described above. Inactiva-tion followed first order kinetics and the inactivation ratewas determined using log linear correlation.

Calculation of Collision Efficiencies. The collision ef-ficiency represents the fraction of collisions (contacts)between suspended bacteriophages and collector soil grainsthat result in attachment. This collision efficiency is describedby the following equation (17):

R) 23

dc

(1- ε)

katt

v1η

(1)

where R is the collision efficiency, dc is the average diameterof the collector (grains), ε is the porosity, katt is the attachmentrate coefficient, v is the average interstitial velocity, and η isthe single collector efficiency. The attachment rate coefficientwas calculated using the equation described by Schijven etal. (10) (see Supporting Information) and the single collectorefficiency η was calculated using the extended multiregres-sion equation of Tufjenki and Elimelech (18) (see SupportingInformation).

ResultsTracer Bromide. To determine the water residence timebetween injection well, two monitoring wells, and abstractionwell, bromide was injected as a conservative tracer. Inaddition to the water residence time, breakthrough char-acteristics of the conservative tracer bromide can be used tocalculate the interstitial flow velocity, dispersivity of theporous medium, and dilution.

The characteristics of bromide breakthrough are pre-sented in Table 1; the breakthrough curves are shown inFigure S2. Maximum breakthrough at the first and secondmonitoring well and the abstraction well was noticed 25, 47,and 72 days after bromide was injected at the injection well,respectively (Table 1). Maximum breakthrough concentrationwas 252 mg L-1 bromide (31.3% of injected concentration)at the first monitoring well, 130 mg L-1 (16.1% of injectedconcentration) at the second monitoring well, and 0.153 mgL-1 (1.89 × 10-4% of injected concentration) at the abstractionwell (Table 1). Because the abstraction well abstractedgroundwater from radial surroundings, the injected bromideplume was diluted with surrounding groundwater at theabstraction well explaining the low bromide concentrationat the abstraction well.

TABLE 1. Breakthrough Characteristics of Bromide and Bacteriophage MS2 and ΦX174 at the Two Monitoring Wells andAbstraction Well

well C0 or Cmaxa (mg L-1; pfu mL-1) ta (days) xa (m) dilution vb (m day-1) rL

b (m) Log C/C0

bromideinjection well 809 0 0 0monitoring well 1 252 25 7.8 3.2 0.33 0.19monitoring well 2 130 47 17.2 6.2 0.38 0.26abstraction well 0.153 72 37.7 5287 0.56 0.23

MS2injection well 6.9 × 109 0 0monitoring well 1 1.4 × 108 18 7.8 -1.7monitoring well 2 5.9 × 106 36 17.2 -3.1abstraction well 4.5 × 102 63 37.7 -7.2

ΦX174injection well 2.5 × 106 0 0monitoring well 1 4.6 × 103 19 7.8 -2.7monitoring well 2 2.8 × 102 37 17.2 -4.0abstraction well 6.9 × 10-3 62 37.7 -8.6a C0 is the concentration injected at the injection well; Cmax is the maximum breakthrough concentration at the

monitoring and abstraction wells; t is the residence time from the injection well; x is the distance to the injection well.b The average interstitial velocity (v) and the longitudinal dispersivity (RL) were determined using the curve-fitting programCXTFIT (19) and were determined for the transport from injection well to each monitoring well or abstraction well.

4590 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 12, 2008

The interstitial velocity and longitudinal dispersivity ofthe water were estimated from the breakthrough curves ofbromide using the curve-fitting program CXTFIT (19). Theinterstitial velocity from injection well to the first or secondmonitoring well was similar, but was higher from injectionwell to abstraction well (Table 1). Apparently, the velocity inthe second section (monitoring well 2 to abstraction well)was slightly higher than in the first section (injection well tomonitoring well 2). The longitudinal dispersivity values werelow and similar for monitoring and abstraction wells (Table1), indicating that the sandy layer of the aquifer was relativelyhomogeneous.

Bacteriophage MS2 and ΦX174. Although it was con-cluded from the bromide tracer experiment that the waterresidence time between injection and abstraction well wasslightly higher than 50-60 days, the bacteriophage data aresufficient to determine the removal of viruses during 50-60days residence time (which is used as length for the microbialprotection zone in many countries) in an anoxic aquifer.

Characteristics of the breakthrough of bacteriophage MS2and ΦX174 are presented in Table 1 and Figure S3. Theresidence time of the bacteriophage MS2 and ΦX174 frominjection well to monitoring and abstraction wells was similar,but lower than the residence time of the conservative tracerbromide (Table 1). This difference in residence time dem-onstrates that MS2 and ΦX174 traveled faster through thesaturated soil than bromide, which has been observed aswell in previous field studies (reviewed in ref (20)). Thisphenomenon was named pore size exclusion and refers tothe assumption that MS2 and ΦX174 did not travel throughthe smallest pore sizes in an aquifer, whereas bromide did.Consequently, a part of the injected bromide in our studytraveled through the smallest pore sizes with the lowestvelocity, whereas bacteriophages traveled through bigger poresizes where higher velocities occurred.

The conservative tracer was only reduced by dispersionand dilution, whereas both bacteriophages were reduced bydispersion, dilution, inactivation, and adsorption. Therefore,C/C0 values of bacteriophages were lower than those forbromide. C/C0 values of ΦX174 were lower than C/C0 valuesof MS2, indicating higher removal of ΦX174 compared toMS2 (Table 1). The log removal of both bacteriophages wasalso plotted against residence time (Figure 1) and distance(Figure S4).The total reduction of bacteriophages was lowfrom injection well to the second monitoring well, and highfrom the second monitoring well to abstraction well (Figure1). The high reduction rate at the abstraction well (7.2 logunits for MS2 and 8.6 log units for ΦX174) was observed withbromide as well (Table 1), and was mainly caused by dilutionof the injected plume at the abstraction well with surrounding

abstracted groundwater. The removal of MS2 or ΦX174 byinactivation and adsorption was estimated by correcting thereduction values for dilution, using the dilution+dispersionfactor of bromide. Removal of bacteriophage ΦX174 byadsorption and inactivation was 4.9 log units, whereas 3.5log units of bacteriophage MS2 were removed (Figure 1).Furthermore, removal of ΦX174 was higher in the first part(injection well to monitoring well 1) and lower in the secondpart (monitoring well 1 to abstraction well), whereas logremoval of MS2 was linear (R2 ) 0.99) over the whole transect(Figure 1).

The amount of removed bacteriophages was split inremoval by inactivation and removal by adsorption. Inac-tivation rates from the survival experiments were used tocalculate removal by inactivation; subsequently total, dilu-tion, and inactivation reduction values were used to calculateremoval by adsorption. Results from these calculationsshowed that 2.3 log units of MS2 were removed by inactivationand 1.2 log units were removed by adsorption. For ΦX174,0.8 log units were removed by inactivation and 4.1 log unitswere removed by adsorption. These calculations showed thatadsorption behavior of MS2 was more conservative, but forΦX174 inactivation was more conservative. Overall, removalof MS2 was lower and thus more conservative than removalof ΦX174 during transport in an anoxic aquifer with arelatively high pH.

Inactivation of Bacteriophage MS2 and ΦX174. Theinactivation of the bacteriophages followed first-order kinet-ics in all samples and inactivation rates were calculated bylog linear regression (Figure S5). The inactivation rate ofΦX174 was lower than the inactivation rate of MS2 (Table2), indicating that ΦX174 is a more conservative virusregarding inactivation than MS2. The inactivation of MS2and ΦX174 did not differ much among the three samplelocations (Table 2). We conclude that inactivation kineticswere similar along the transect where bacteriophages movedthrough the aquifer and that there were no subpopulationsof phages in relation to inactivation.

The inactivation of attached bacteriophages was deter-mined as described by Schijven et al. (10), which is bycalculating the tail slope of the breakthrough curves of MS2and ΦX174 at logarithmic scale. Inactivation rates of attachedMS2 or ΦX174 phages were comparable with inactivationrates of free MS2 or ΦX174 phages (Table 2), demonstratingthat inactivation behavior was comparable for attached andfree bacteriophages. Only the inactivation rate of attachedMS2 determined from the tail of the breakthrough curve atthe second monitoring well and of attached ΦX174 deter-mined from the curve at the injection well showed highervalues (Table 2). However, these high values were less reliable,since regression coefficients of the tail slope were lower (R2

is 0.92 or 0.84) compared to regression coefficients of theother tail slopes (R2 > 0.95).

Collision Efficiencies. The collision efficiency holdsinformation about the adsorption capacity of phages to soilparticles. The collision efficiency of MS2 or ΦX174 was similarat each monitoring or abstraction well (Table 2). Thus,adsorption of MS2 and ΦX174 was constant over the wholetransect from injection to abstraction well, which indicatesthat available attachment sites were homogenously distrib-uted throughout the aquifer.

The collision efficiency of bacteriophage MS2 at each wellwas 3.5-5 times lower than the collision efficiency of ΦX174(Table 2). Consequently, we conclude that adsorptionbehavior of phage MS2 was more conservative than that ofΦX174. This can be explained by the lower iso-electric pointof MS2 (21) resulting in a more negatively charged phageand a higher repulsion by the negatively charged soil particlesthan the less negatively charged ΦX174 phages (21).

FIGURE 1. Removal of bacteriophage MS2 and ΦX174 versusresidence time. Symbols: b, MS2 removal; 9, ΦX174 removal; (MS2 removal without dilution; 2, ΦX174 removal withoutdilution; - - - is the linear regression line for the removal ofbacteriophage MS2.

VOL. 42, NO. 12, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 4591

DiscussionAdsorption and Inactivation. Removal of bacteriophages inthe saturated zone of sandy aquifers has been studied atdifferent field locations (6–8, 10–12). At these other locations,high removal rates of both bacteriophage MS2 and ΦX174were observed during relatively short residence times in theaquifer. The low removal rates of MS2 and ΦX174 observedin our study were clearly in contrast with results from theseother studies (Figure 2). It is well-known that grain size andgroundwater pH can impact removal of phages (reviewed inrefs 21, 22). Grain sizes of sand and groundwater pH in thestudied aquifers were in most cases similar to grain sizes andpH at the field location used in our study. Hence, grain sizeand pH were not responsible for the observed differencebetween our study and other field studies. Five of the sixstudies that investigated removal of bacteriophages MS2 and/or ΦX174 were done in aquifers that contained oxic ground-water (6–8, 10, 12) and one study investigated removal ofMS2 in an anoxic aquifer where oxic surface water wasinfiltrated, resulting in oxic conditions around the infiltrationsite (11). Consequently, it seems obvious that the anoxic redox

condition in the aquifer used in our study was the apt causefor the observed lower removal rates. The conclusion thatanoxic conditions in the aquifer were responsible for the lowremoval rate of MS2 and ΦX174 was supported by laboratorystudies. It has been observed that inactivation of viruses ingroundwater was lower under anoxic conditions (23), andcolumn studies demonstrated that the presence of oxidizedmetal ions like ferric oxihydroxides, which are present underoxic conditions but are likely to be absent under anoxicconditions, resulted in higher virus adsorption (24–26). Inconclusion, the anoxic aquifer with a relatively high pH andlow organic content we used in our transport study hasunfavorable conditions for virus removal and can be seen asa worst case sandy aquifer.

The seeding concentration of bacteriophage MS2 was30,000 times higher than that of bacteriophage ΦX174. In aprevious study, it was demonstrated that the removal rateof MS2 in slow sand filters was not affected when the seedingconcentration of MS2 increased 1,000 times (27). Conse-quently, the difference in seeding concentration of MS2 andΦX174 in our study is unlikely to be the cause for the observeddifference in the removal rate of MS2 and ΦX174. As statedearlier, the lower iso-electric point of MS2 compared to ΦX174(21) results in a more negatively charged MS2 than ΦX174and consequently in a higher repulsion by the negativelycharged soil particles. In conclusion, the difference in surfacecharge is the apt cause for the different removal rates of MS2and ΦX174 observed in our study.

Inactivation rates obtained for MS2 and ΦX174 at 5-13°C in other studies were generally higher than inactivationrates obtained in our study (7, 10, 11, 28–31). The lowerinactivation rates observed in our study confirm a previouslaboratory study that demonstrated that anoxic conditionsresulted in lower inactivation rates (23). Collision efficienciesof MS2 and ΦX174 have been calculated in other field studiesas well, but these studies were mainly performed in oxicaquifers (6, 9–11). The collision efficiencies obtained in thosestudies were 10-100 times higher than the collision ef-ficiencies observed in our study, indicating that adsorptionof bacteriophages under oxic conditions was higher thanthat under anoxic conditions. A similar low collision efficiencywas reported for MS2 during transport in the anoxic part ofan aquifer that changed from anoxic to oxic during infiltrationof oxic surface water (11, 13), supporting our observationthat transport in an anoxic aquifer results in low collisionefficiencies.

In conclusion, the removal rate of bacteriophages MS2and ΦX174 in an anoxic aquifer was low compared to removalrates of both phages in oxic aquifers. The low removal ratein anoxic aquifers was caused by a lower inactivation rateand a lower adsorption rate of both phages at anoxicconditions.

Surrogates for Human Pathogenic Viruses. We have usedbacteriophages MS2 and ΦX174 as surrogates for humanpathogenic viruses. Compared to human pathogenic viruses

TABLE 2. Inactivation Rates and Collision Efficiency of Bacteriophage MS 2 and ΦX174

bacteriophage MS2 bacteriophage ΦX174

inactivation (log day-1) inactivation (log day-1)

locationa free attached collision efficiency free attached collision efficiency

seeding tank 0.031 0.038 0.008 0.020monitoring well 1 0.040 0.036 3.4 × 10-5 0.013 0.017 1.7 × 10-4

monitoring well 2 0.037 0.055 2.9 × 10-5 0.018 0.019 1.1 × 10-4

abstraction well 2.8 × 10-5 1.0 × 10-4

average 0.036 0.043 3.0 × 10-5 0.013 0.019 1.3 × 10-4

a For the inactivation of free bacteriophages samples were taken from the seeding tank, monitoring well 1 duringbreakthrough, and monitoring well 2 43 days after maximum breakthrough.

FIGURE 2. Removal of MS2 (A) and ΦX174 (B) observed atdifferent field locations. Symbols: 9, this study; [ ref (8); b, ref(10); 2, ref (11); 4, ref (7); 0, ref (6); O, ref (12).

4592 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 12, 2008

the adsorption kinetics of MS2 were more conservative andcan be considered worst case when soil organic content waslow (6, 12, 21, 32), as was the case at our field site (organiccontent ∼0.15%).

The inactivation rate of bacteriophage ΦX174 was lowerthan that of MS2. Hence, the inactivation rate of MS2 is notworst case compared to other bacteriophages. Therefore,the inactivation rate of MS2 obtained in our study wascompared with inactivation rates published for humanpathogenic viruses (33). Inactivation rates of human patho-genic viruses varied between 0.0043 and 0.52 log day-1. Thelow inactivation rates reported for human pathogenic virusesare lower than the inactivation rate for MS2 observed in ourstudy. However, low inactivation rates for human pathogenicviruses were obtained in studies where groundwater was notincubated under the environmental conditions of the aquifer(34), where inactivation did not follow first-order kinetics(35), or where incubation of viruses without host cells showedan increase in virus numbers (36, 37). The low inactivationrates for a number of pathogenic viruses were reported onlyonce, mostly in a study that has not been published in apeer-reviewed journal (35). It can not be deduced from thatstudy whether the groundwater incubations were undersimilar conditions as observed in the aquifer (35). Moreover,a higher inactivation was observed during the first 60 days,followed by a lower inactivation over the next 200 days (35).Consequently, the published low inactivation rates for humanpathogenic viruses seems unreliable compared to the higherpublished inactivation rates, which were equal to or higherthan the inactivation rate obtained for MS2 in our study. Asa result, we conclude that the inactivation rate observed forMS2 in our study could be considered worst case when relatedto human pathogenic viruses, as was concluded before (21).

Microbial Protection Zone. In previous field studies, 8log removal of bacteriophages was achieved within 30 daysresidence time (6–8, 10–12). As a result, a microbial protectionzone based on 50-60 days residence time seemed more thansufficient to protect the abstraction well against virusbreakthrough from a contamination source. However, weobserved only 3.5 log removal of MS2 during 63 days residencetime and it was unclear if such low removal rates wouldprotect the abstraction well against virus breakthrough.Therefore, the required length of the protection zone in caseof a worst case scenario and Dutch legislation was calculatedusing data from our field study. The highest risk of con-tamination of groundwater with human pathogenic virusesis related to a leaking sewer in the vicinity of the abstractionwell. In the Dutch water decree, it is stated that drinkingwater should not exceed an infection risk of 1 out of 10,000persons per year (15). Because groundwater has a highhygienic quality, there are no treatment steps to removepathogenic microorganisms from groundwater, nor is drink-ing water chemically disinfected in The Netherlands. Con-sequently, abstracted groundwater should have a virusconcentration below 1.2 × 10-6 N L-1 to remain below theinfection risk of 10-4 persons per year (38). This means thatthe virus concentration should be reduced below 1.2 × 10-6

N L-1 during transport in the aquifer from a leaking sewerto abstraction well. The highest Entero- and Reovirusconcentration in raw sewage from a sewage plant inApeldoorn, The Netherlands was 833 and 2143 N L-1,respectively (39). Thus, during transport from leaking sewerto abstraction well in an anoxic aquifer, 8.8 log removal ofEnterovirus and 9.3 log removal of Reovirus should beachieved.

Viruses that move from leaking sewer to abstraction wellare diminished by inactivation, adsorption, and dilution.Schijven et al. (13) assumed in their modeling studies thata leaking rate of 1 m3 day-1 would be a realistic value forsewage leaks in The Netherlands. An abstraction rate of 1000

m3 day-1 is obtained by a small groundwater well that is usedfor drinking water production in The Netherlands. A leakingrate of 1 m3 day-1 and an abstraction rate of 1000 m3 day-1

result in 3 log reduction caused by dilution. Consequently,5.8 log units of Enterovirus and 6.3 log units of Reovirus haveto be removed by inactivation and adsorption. The removalrate of Reovirus and Enterovirus during transport in an anoxicaquifer is unknown. Since the removal of bacteriophage MScan be considered as a surrogate for conservative virustransport (reviewed in (21) and observations from this study),we used MS2 removal obtained in our study to calculateremoval of Reovirus and Enterovirus during transport in ananoxic aquifer. We demonstrated in our study that removal(inactivation and adsorption) of MS2 in an anoxic aquiferwith a relatively high pH was log linear with residence timeand can be described by: log (C/C0) ) 0.0576 × t, where log(C/C0) is log removal and t is residence time in the aquifer.Using this equation, a 5.8 log removal of Enterovirus will beobtained with a residence time of 101 days, and a 6.3 logremoval of Reovirus with a residence time of 109 days. Thus,the currently used guideline in The Netherlands for the lengthof the microbial protection zone (based on 60 days residencetime) may not protect the abstraction well sufficiently undera worst case scenario of a sewer leaking directly in a shallowanoxic aquifer with a relatively high pH. However, the lengthof the protection zone should not have a residence time of1-2 years as was proposed in studies that used a virustransport model to calculate the length of the microbialprotection zones around Dutch anoxic aquifers (13, 14). Thereason for the discrepancy between our field study and themodeling study is that values of the most sensitive modelparameters, inactivation and collision efficiency, were un-realistically low in the modeling studies (13, 14).

Overall, our study demonstrated that removal of virusesduring transport in an anoxic aquifer was considerably lowerthan that in oxic aquifers. The low removal of viruses in ananoxic aquifer was caused by a lower inactivation andadsorption rate under anoxic conditions. Moreover, weconclude from our study that, under a worst case scenarioof a leaking sewer, the currently used microbial protectionzones in some countries, which are based on 50-60 daysresidence time (1, 2), may not be sufficient to producedrinking water that meets an infection risk of 10-4 personsper year. Consequently, the currently used length of microbialprotection zones around shallow abstraction wells in anoxicsandy aquifers should be reconsidered and the total traveltime between contamination sources and abstraction wellshould preferably be extended to 110 days.

AcknowledgmentsThis study was financed by the Dutch water supply companiesas part of the joint research program (BTO). We thank KeesMaas and Gijsbert Cirkel for help on the geohydrology andAnke Brouwer and Anita Lugtenberg for technical assistance.

Supporting Information AvailableDetailed description of the collision efficiency calculations,a schematic representation of the aquifer, breakthroughcurves of bromide, MS2, and ΦX174 at the monitoring andabstraction wells, inactivation curves of free MS2 and ΦX174,and removal of MS2 and ΦX174 as function of distance. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited(1) Chave, P.; Howard, G.; Schijven, J. F.; Appleyard, S.; Fladerer,

F.; Schimon, W. Groundwater protection zones. In ProtectingGroundwater for Health; Schmoll, O., Howard, G., Chilton, J.,Chorus, J., Eds.; World Health Organization: Cornwall, UK, 2006;pp 465-492.

VOL. 42, NO. 12, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 4593

(2) Commissie Bescherming Waterwingebieden. Guidelines andrecommendations for the protection of groundwater intake areas;VEWIN-RID: Rijswijk, The Netherlands, 1980.

(3) Knorr, M. Die Schutzzonenfrage in der trinkwasser-hygiene.Das Gas- Wasserfach 1937, 80, 350–355.

(4) Ferguson, C.; de Roda Husman, A. M.; Altavilla, N.; Deere, D.;Ashbolt, N. Fate and transport of surface water pathogens inwatersheds. Crit. Rev. Environ. Sci. Technol. 2003, 33, 299–361.

(5) Robertson, J. B.; Edberg, S. C. Natural protection of spring andwell drinking water against surface microbial contamination.I. Hydrogeological parameters. Crit. Rev. Microbiol. 1997, 23,143–178.

(6) DeBorde, D. C.; Woessner, W. W.; Kiley, Q. T.; Ball, P. Rapidtransport of viruses in a floodplain aquifer. Water Res. 1999, 33,2229–2238.

(7) DeBorde, D. C.; Woessner, W. W.; Lauerman, B.; Ball, P. N.Virus occurrence and transport in a school septic system andunconfined aquifer. Ground Water 1998, 36, 825–834.

(8) Medema, G. J.; Stuyfzand, P. J. Removal of microorganisms uponbasin recharge, deep well injection and river bank filtration inthe Netherlands. In Proceedings of the 4th InternationalSymposium on Artificial Recharge, Adelaide, Australia, Sep 22-26, 2002.

(9) Pang, L.; Close, M.; Goltz, M.; Noonan, M.; Sinton, L. Filtrationand transport of Bacillus subtilis spores and the F-RNA phageMS2 in a coarse alluvial gravel aquifer: Implications in theestimation of setback distances. J. Contam. Hydrol. 2005, 77,165–194.

(10) Schijven, J. F.; Hoogenboezem, W.; Hassanizadeh, S. M.; Peters,J. H. Modeling removal of bacteriophage MS2 and PRD1 bydune recharge at Castricum, the Netherlands. Water Resour.Res. 1999, 35, 1101–1111.

(11) Schijven, J. F.; Medema, G. J.; Vogelaar, A. J.; Hassanizadeh,S. M. Removal of microorganisms by deep well injection. J.Contam. Hydrol. 2000, 44, 301–327.

(12) Woessner, W. W.; Ball, P. N.; DeBorde, D. C.; Troy, T. L. Viraltransport in a sand and gravel aquifer under field pumpingconditions. Ground Water 2001, 39, 886–894.

(13) Schijven, J. F.; Mulschlegel, J. H.; Hassanizadeh, S. M.; Teunis,P. F.; de Roda Husman, A. M. Determination of protection zonesfor Dutch groundwater wells against virus contamination. J.Water Health 2006, 4, 297–312.

(14) van der Wielen, P. W. J. J.; Blokker, M.; Medema, G. J. Modelingthe length of microbiological protection zones around phreaticsandy aquifers in The Netherlands. Water Sci. Technol. 2006,54 (3), 63–69.

(15) Dutch Government. Waterleidingbesluit 2001; The Hague: TheNetherlands, 2001.

(16) Stuyfzand, P. J. Hydrochemistry and hydrology of the coastaldune area of the Western Netherlands. PhD thesis, Nieuwegein,The Netherlands, 1993.

(17) Yao, K. M.; Habibian, M. T.; O’Melia, C. R. Water and wastewater filtration: concepts and applications. Environ. Sci. Technol.1971, 5, 1105–1112.

(18) Tufenkji, N.; Elimelech, M. Correlation equation for predictingsingle-collector efficiency in physicochemical filtration insaturated porous media. Environ. Sci. Technol. 2004, 38, 529–536.

(19) Toride, N.; Leij, F. J.; van Genuchten, M. T. The CXTFIT code forestimating transport parameters from laboratory or field tracerexperiments. Version 2.0; U.S. Department of Agriculture:Riverside, CA, 1995; Report No. 137.

(20) Taylor, R.; Cronin, A.; Pedley, S.; Barker, J.; Atkinson, T. Theimplications of groundwater velocity variations on microbial

transport and wellhead protection - review of field evidence.FEMS Microbiol. Ecol. 2004, 49, 17–26.

(21) Schijven, J. F.; Hassanizadeh, S. M. Removal of viruses by soilpassage: overview of modeling, processes and parameters. Crit.Rev. Environ. Sci. Technol. 2000, 31, 49–125.

(22) Jin, Y.; Flury, M. Fate and transport of viruses in porous media.Adv. Agron. 2002, 77, 39–102.

(23) Gordon, C.; Toze, S. Influence of groundwater characteristicson the survival of enteric viruses. J. Appl. Microbiol. 2003, 95,536–544.

(24) Abudalo, R. A.; Bogatsu, Y. G.; Ryan, J. N.; Harvey, R. W.; Metge,D. W.; Elimelech, M. Effect of ferric oxyhydroxide grain coatingson the transport of bacteriophage PRD1 and Cryptosporidiumparvum oocysts in saturated porous media. Environ. Sci.Technol. 2005, 39, 6412–6419.

(25) Loveland, J. P.; Ryan, J. N.; Amy, G. L.; Harvey, R. W. Thereversibility of virus attachment to mineral surfaces. ColloidsSurf. A 1996, 107, 205–221.

(26) Schijven, J. F.; Hassanizadeh, S. M.; de Bruin, H. A. Columnexperiments to study nonlinear removal of bacteriophages bypassage through saturated dune sand. J. Contam. Hydrol. 2002,58, 243–259.

(27) Hijnen, W. A. M.; Schijven, J. F.; Bonne, P.; Visser, A.; Medema,G. J. Elimination of viruses, bacteria and protozoan oocysts byslow sand filtration. Water Sci. Technol. 2004, 50 (1), 147–154.

(28) Blanc, R.; Nasser, A. Effect of effluent quality and temperatureon the persistence of viruses in soil. Water Sci. Technol. 1996,33 (10-11), 237–242.

(29) Yahya, M. T.; Galsomies, L.; Gerba, C. P.; Bales, R. C. Survivalof bacteriophages MS2 and PRD1 in ground water. Water Sci.Technol. 1993, 27, 409–412.

(30) Yates, M. V.; Gerba, C. P.; Kelley, L. M. Virus persistence ingroundwater. Appl. Environ. Microbiol. 1985, 49, 778–781.

(31) Yates, M. V.; Stetzenbach, L. D.; Gerba, C. P.; Sinclair, N. A. Theeffect of indigenous bacteria on virus removal in ground water.J. Environ. Sci. Health, A 1990, 25, 81–100.

(32) Goyal, S. M.; Gerba, C. P. Comparative adsorption of humanenteroviruses, simian rotavirus, and selected bacteriophagesto soils. Appl. Environ. Microbiol. 1979, 38, 241–247.

(33) Pedley, S.; Yates, M.; Schijven, J. F.; West, J.; Howard, G.; Barrett,M. Pathogens: Health relevance, transport and attenuation. InProtecting Groundwater for Health; Schmoll, O., Howard, G.,Chilton, J., Chorus, J., Eds.; World Health Organization: Cornwall,UK, 2006; pp 49-80.

(34) Jansons, J.; Edmonds, L. W.; Speight, B.; Bucens, M. R. Survivalof viruses in groundwater. Water Res. 1989, 3, 301–306.

(35) Dizer, H.; Filip, Z.; Lopez, J. M.; Milde, G.; Nasser, A.; Seidel, K.Laborversuche zur persistenz und zum transportverhalten vonviren. Umweltbundesamt Materialien 1985, 2/85, 20–26.

(36) Nasser, A. M.; Oman, S. D. Quantitative assessment of theinactivation of pathogenic and indicator viruses in natural watersources. Water Res. 1999, 33, 1748–1752.

(37) Nasser, A. M.; Tchorch, Y.; Fattal, B. Comparative survival of E.coli, F+bacteriophages, HAV and Poliovirus 1 in wastewaterand groundwater. Water. Sci. Technol. 1993, 27 (3-4), 401–407.

(38) de Roda Husman, A. M.; Medema, G. J.; Teunis, P. F. M.; Schijven,J. F. Inspection guideline analysis microbiological safety water;VROM-inspection; The Hague: The Netherlands, 2004.

(39) Lodder, W. J.; de Roda Husman, A. M. Presence of Norovirusesand other enteric viruses in sewage and surface waters in theNetherlands. Appl. Environ. Microbiol. 2005, 71, 1453–1461.

ES800156C

4594 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 12, 2008