Environ. Sci. Technol. 2011, 45, 636–642
Analytical Solution for the Modeling of the Natural Time-Dependent Reduction of Waterborne Viruses Injected into Fractured Aquifers C O S T A N T I N O M A S C I O P I N T O , * ,† ROSANNA LA MANTIA,† CATERINA LEVANTESI,‡ VALTER TANDOI,‡ MAURIZIO DIVIZIA,§ DOMENICA DONIA,§ ROSANNA GABRIELI,§ AND ANNA RITA PETRINCA§ National Research Council, Water Research Institute, via Francesco De Blasio, 5 -70123 Bari, Italy, National Research Council, Water Research Institute, via Salaria km. 29, 300 - 00016 Monterotondo, Roma, Italy, and Department of Public Health, Faculty of Medicine, Hygiene, Tor Vergata University, via Montpellier, 1- 00133-Rome, Italy
Received July 16, 2010. Revised manuscript received November 26, 2010. Accepted December 1, 2010.
We propose an analytical solution in order to explain the processes that determine the fate and behavior of the viruses during transport in a fractured aquifer at Salento (Italy). The calculations yield the efficiency of filtration in fractures at a site near Nardo` (Southern Italy) in reducing the numbers of enteric viruses (i.e., Enteroviruses and Norovirus) in secondary municipal effluents that have been injected in the aquifer over the period 2006-2007. The model predicted, by a theoretical expression, the time-dependent rate of virus reduction, which was in good agreement with field data. The analytical solution yields the achievable “Log reduction credits” (1) for virus reduction in wells located at the setback distances that are usually adopted in local drinking water regulations. The resulting new analytical formula for the time-dependent reduction of viruses during subsurface transport can easily be applied in health risk-based models used to forecast the spread of waterborne diseases and provides appropriate criteria (i.e., distances) needed to meet standards for the quality of drinking water derived from undisinfected groundwater.
Introduction Groundwaters form the main sources of water in most semiarid zones around the world, even though their disinfection for drinking is not always practiced, particularly in underdeveloped countries. Much of the groundwater used for drinking in the U.S. is also not disinfected (1). A great amount of potable water in the world is derived from water reservoirs in fractured limestone, and ∼1.6 billion people depend upon the health of carbonate terrains and aquifers for their water supply (2). In the Salento region (southern Italy) the carbonate fractured aquifer supplies 126 million * Corresponding author e-mail: ba.irsa.cnr.it. † Water Research Institute, Bari. ‡ Water Research Institute, Roma. § Tor Vergata University. 636
9
costantino.masciopinto@
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 2, 2011
m3/yr of drinking water, i.e. 80% of the drinking water requirement of the population (∼800,000 inhabitants); here the presence of numerous houses scattered throughout the peninsula which rely on unauthorized wells that supply undisinfected water is a well-known problem. Some concern was expressed in 1998 (3) regarding groundwater quality and public health in the Salento area. Enteric viruses, which include the Rotaviruses, Hepatitis A virus (HAV), Norwalk, Adenoviruses, Enteroviruses, and others, can be transmitted via a faecal-oral route. Giardia and Cryptosporidium are the most common parasitic protozoa found in groundwaters of industrialized countries, which may lead to waterborne outbreaks of giardiasis and cryptosporidiosis. Due to their persistence in groundwater, pathogens may be found some distance away from the source of the contamination and even in tap water (4). An interesting investigation has been proposed by Taylor et al. (5), which highlights that field studies of sewage contaminations provide evidence of high bacteria and virus counts in groundwater flow paths that permit the transport of microorganisms by rapid or extreme water velocities. This means that fractured rocks, which have preferential rapid flow pathways with respect to porous granular media, are more vulnerable to the groundwater contamination. In light of this, models (6, 7) can be used to explore the total reduction of pathogens by filtration and inactivation in fractures. Simulations may establish the extent to which this removal is capable of mitigating the possible risks posed by pathogens in undisinfected drinking water. Field experiments carried out by Scheibe et al. (8) have observed that the apparent rate of bacterial attachment, particularly as parametrized by the collision efficiency in filtration-based models, decreases with transport distance (i.e., exhibits scaledependency). Furthermore, other known models (9, 10) that consider a time-dependent variability of the rate of virus reduction have been derived from the statistical elaboration of experimental measurements (i.e., based on empirical assumptions). This means that at a given location, after the maximum virus deposition is achieved, the attachment may decrease with time as a result of the decrease in the natural filtration efficiency in the elimination of the virus from the groundwater. The objective of this study is to investigate how the fate and transport of viruses and protozoa in fractures may affect undisinfected groundwater in wells, such as at the Nardo` site. Successful simulations are achieved by accommodating a new theoretical formulation for the temporal variation of the virus reduction rate in the transport equation. The proposed time-dependent rate of virus reduction, which is i) site-specific and ii) dependent upon the inactivation constants of the suspended/deposited virus under consideration, was obtained by solving the virus transport equations along a generic pathway using a Lagrangian approach (11, 12). Simulation results were supported by field and laboratory filtration tests using breakthrough curves of somatic coliphages.
Material and Methods Study Area. In the Salento peninsula (Figure 1), groundwater flow occurs at a low pressure within fractured limestone (99% of CaCO3, silica, iron, and aluminum oxides) (13) and is affected by withdrawals for use in drinking and irrigation. Hydrological studies at the Nardo` site (6) have shown that the groundwater has a variable saturated thickness that has an average value of 30 m and that the water flows along preferential horizontal pathways under a head gradient of 10.1021/es102412z
2011 American Chemical Society
Published on Web 12/20/2010
FIGURE 1. Location of the sampling points (SP1, SP2, SP3, SP4, and SP5) during 2006-2007 monitoring at the Nardo` site and modeled (5) groundwater contour heads (in m above sea level) during winter 2007. 3.75 × 10 -4 m/m. The water table at the injection site (sinkhole) lies approximately 32 m below the surface. The water quality of both the groundwater and injectant together with the injectant flow rate are summarized in Table 1. The Nardo` aquifer was monitored seasonally by sampling four times during 2006 and three times during 2007, at five sampling points (SP): SP1-wastewater treatment plant (WWTP), SP2-injection, SP3-well, SP4-well, and SP5-well (see Figure 1). Sampling point SP1 was taken in the secondary effluent stream of a municipal plant after treatment by an activated sludge process. Sampling point SP2 was the aquifer injection point (i.e., the sinkhole). Wells SP3 and SP4 are located 320 and 500 m, respectively, from the injection site and downgradient with respect to the groundwater flow. The well SP5, located 1000 m from the injection site, was upstream of it, with respect to the main direction of groundwater flow (i.e., without injection). Sample Collection and Microbiological Methods. Groundwater was sampled during seven campaigns in different seasons of 2006 (January, April, July, and November) and 2007 (February, May, and September) at the same sampling locations. Large volumes of 5-120 L and 200-250 L were sampled for the protozoa (14) and virus analyses, respectively. The groundwater was filtered directly on site by pumping from wells into electropositive cartridges of external diameter 5 cm and length 25 cm (cod. ZPMK CUNO, Meridian, USA) for viruses, while protozoa were retained using IDEXX FiltaMax foam filters. The cartridges and filters were analyzed at Rome IRSA and University Laboratories, while pathogen indicators, i.e. bacterial count, E. coli, coliforms (15), En-
terococci (16), somatic coliphages (14), and clostridium spores (17) were analyzed at Bari IRSA Laboratory. Filter modules were eluted using IDEXX Filta-Max xpress and the eluate was concentrated to 10 mL by centrifugation (15 min, 2000 rpm). Cysts and oocysts were purified by means of immuno-magnetic separation (Dynabeads Crypto-Combo) and stained with fluorescently labeled monoclonal antibodies (MERIFLUOR Crypto & Giardia kit, Meridian Bioscience, Europe) and with DAPI (0.4 µg/mL of phosphate buffered saline (PBS) at 37 °C for 15 min). Giardia and Cryptosporidium cysts and oocysts were finally enumerated by epifluorescence microscopy (Olympus BX-51) and phase contrast observation. Enterovirus (Poliovirus, Echovirus, Coxsackievirus), Hepatitis A virus (HAV), and Norovirus genotypes I (GGI) and II (GGII) (18) were tested for using reverse transcription polymerase chain reactions (RT-PCRs) on cartridge eluate obtained using beef extract (3%). Eluates were treated with PEG 6000 (10% w/v final concentration), agitated overnight at +4 °C, and concentrated by centrifugation (8000 rpm for 1 h at +4 °C) to remove RT-PCR inhibitors. After centrifugation the pellet was suspended in PBS solution (after elimination of calcium and magnesium). The samples were then treated with chloroform (30% v/v) for lipids degradation and were ultracentrifuged at 35000 rpm (1 h at 4 °C). The pellet was finally suspended in PBS (depleted of calcium and magnesium) and stored at -20 °C. Viral RNA was extracted using TRIzol LS Reagent (invitrogen by life technologies). Enterovirus and HAV cDNA was synthesized by reverse transcription, using specific primers, and successively amplified by means of two PCR reactions that produced 155 base pair (bp) amplified products for Enteroviruses and 317 bp amplified products for HAV. Norovirus GGI and GGII were analyzed by means of RT-seminested PCR using specific primers that targeted highly conserved genomic regions (the capsidic region for GGI and the polymerasic region for GGII). GG I and GG II Noroviruses were analyzed separately using specific primers (19), as reported by Gabrieli et al. (18). The specificity of the RT-PCR reaction in the positive samples was confirmed by sequencing the obtained amplified DNA sequences. Sequence data were analyzed and compared to collections of data available online (http://www.ncbi.nlm.nih.gov/). The software package ClustalW2 (www.ebi.ac.uk) was used for the alignment. Water samples that tested positive for Enterovirus and Norovirus using RT-PCR were also tested for culturability on BGM (buffalo green monkey kidney) cells, by observing cell monolayers for cytopathic effects over a 15-day period. Quantitative Virus Analysis. The presence of Enterovirus and Norovirus in the positive samples was determined using real-time RT-PCR with SYBR Green stains. An Enterovirus calibration curve was obtained using serial dilution (from
TABLE 1. Microbiological Parameters of the Injected Water and Groundwater at the Nardo` Site
3
total flow rate (m /d) electrical conductance (mS/cm) pH suspended solids (mg/L) bacterial count (37 °C CFU/mL) total coliforms (MPN/100 mL) E. coli (MPN/100 mL) Enterococci (CFU/100 mL) sulfite-reducing clostridium spores (CFU/100 mL) somatic coliphagesb (PFU/100 mL) a
groundwatera
average bias (%)
galatone WWTPa effluent
injected watera mean
320 m
500 m
1000 m
(15 (2 (0.01 (33 (3 (68 (72 (29 (8
1300 1.7 7.0 30 24886 42814 406 6083 10408
14500 1.29 7.9 8.4 27950 41901 192 3094 913
SP3 1.17 7.4 9.9 11004 15635 14 539 1845
SP4 0.97 7.6 3.0 7457 8257 66 1115 87
SP5 1.18 7.4 2.9 305 1129 9 161 173
(22
9492
6168
2061
134
131
Mean values obtained from six samples taken between January 2006 and May 2007.
b
Detection limit: 0.12 PFU/100 mL.
VOL. 45, NO. 2, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
637
FIGURE 2. Sketch of a virus particle (dm) trajectory in a 2D single fracture of a limestone aquifer and originating from position x0; x(t; a) is the function of particle location (a) in the fracture plane at current time t; dm is the initial amount of massparticle with concentration C0 released at t ) 0 in the inlet section (sinkhole). 105 to 101 genomic copies (gc) /µL) of a commercially available Enterovirus RNA standard (Armored RNA ENTEROVIRUS). Armored RNA Norwalk virus GG I (from 2.4 × 105 to 2.4 × 101 gc/µL) and Armored RNA Norwalk virus GG II (from 2.1 × 105 to 2.1 × 101 gc/µL) were used instead for the Norwalk virus calibration curves. The viral genomic region ORF 1 (RNA polymerase) was targeted for both Norovirus genomic groups. One-step RT-PCR was performed and SYBR Green stain was used to follow the amplification curves, while the specificity of the amplification was confirmed using melting curve analysis.
Theoretical Developments Colloids generally range in size from 1 nm to 10 µm. Biocolloids, such as bacteria, range in size from 0.2 to 5 µm and protozoa from 8 to 12 µm (21), whereas viruses are at the lower end of the colloid size distribution, ranging in size from 0.02 to 0.3 µm. At the test site the fractured aquifer was discretized in a 3D set of 23 ( 3 parallel fractures (Figure 2). According to the colloid filtration theory (20), two-dimensional virus transport with first-order attachment and inactivation in each single fracture with variable aperture can be derived from the mass balance of the virus numerical particles suspended in the aqueous phase and subject to elimination due to deposition on the fracture walls and inactivation 2 ∂C* 2 dC[t, x(t)] )- λC - λ*C* dt b ∂t b
(1)
In eq 1 C [M/L3] is the concentration of the suspended virus, C* [M/L2] is the concentration of the virus deposited on the fracture surfaces along a single trajectory, expressed as a mass of viruses per unit area of the fracture surface, b [L] is the local aperture in the bedding plane fracture, and λ [t-1] and λ* [t-1] are the inactivation constants for suspended and attached viable (or infective) virus particles, respectively. Note that eq 1 was derived using the Lagrangian approach. 638
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 2, 2011
Indeed, the continuous virus injection at t0 ) 0 over the groundwater volume in the inlet section x ) x0 (or sinkhole) can be subdivided in the initial amount of particles with mass (see Figure 2) dm ) C0(x0)dx, while the change in the attached virus count on each fracture surface with inactivation may be described by the following generalized mass balance expression C ∂C∗ ) rf - rrC∗ - λ∗C∗ ∂t b
(2)
where rr [t-1] is the virus detachment rate coefficient, and rf [L2/t] is the attachment rate coefficient, which can be defined as (21) rf ) κU(x)F(C∗)
(3)
where U(x) is the steady-state local water velocity [L/t] in the single fracture, κ [L] is the virus deposition coefficient, and F(C*) (dimensionless) is the value of the dynamic blocking function (DBF) (21, 22), which accounts for the portion of the fracture surface that is available for virus attachments (the surface exclusion effect). The DBF ranges from 1 for a fracture surface free of virus (C* ) 0) to 0, for a fracture surface completely covered (monolayer) by deposited viruses. The deposition coefficient depends on several physicochemical properties of virus particles and fractures as well as on microscopic virus deposition mechanisms (i.e., Brownian diffusion, van der Waals and electrostatic forces) (i.e., collector efficiency) (12, 23) and on the shape, size, and physiological state (9) of the virus under consideration (i.e., collision efficiency) (8). The initial (t ) 0) conditions for a continuous injection may be written as x(0) ) x0
(4a)
C(0, x0) ) C0
(4b)
C(0, x) ) 0
(4c)
C*(0, x0) ) 0
(4d)
The expected concentration at current time t, in a single fracture can be determined by solving eqs 1-4a as functions of particle pathways and total number of collected particles at the observation wells. The integration of equations 1-4a leads to the following solution (see Supporting Information (SI) for details) ln
[
]
C(t, x, y) ) -(A1 - A2)t C0
(5)
where the coefficients A1 (t-1) and A2(t-1) may be set as A1 )
(
2 rf + λ b2
)
(6)
and A2 ) -
2 1 rr exp[-(rr + λ*)τ] 2 r f r + λ* b r
(7)
and must be estimated before integrating eq 5, using predetermined values of time τ [t] that account for the time required by numeric particles (i.e., viruses) released at the injection site to arrive at the observation well. The timedependent rate coefficient may be defined as λj ) (A1 - A2)
(8)
The left-hand side of eq 8 is always positive. The value of A1 depends on the inactivation constant λ and the attachment virus coefficient rf, while A2 is a function of the attachment/ detachment rate coefficients (rf and rr) and of the inactivation constant λ*. The time-dependent reduction rate can thus be determined after the site-specific coefficients rf and rr and the inactivation constants (λ and λ*) have been assigned by fitting the experimental field data collected for the virus under consideration. Moreover, the rate of virus reduction is a function of both virus inactivation and deposition and is delayed by detachment effects. The latter, by increasing with time, decrease the deposition and will control the rate coefficient of the virus reduction. Two separate codes (24) were used at the Nardo` site to solve the 2D-groundwater flow problem in a set of parallel fractures (5) (see Figure 2) and to determine the breakthrough curves of the virus count at the observation wells, respectively.
Model Test In order to support the results yielded by the model, both field and laboratory filtration tests were carried out to investigate the reduction in coliphages over more than 30 days of filtration. At the Nardo` site two specific tests were carried out using frequent sampling in well SP4, just after recharge operations were started again following periodic (twice per year, during May and November) maintenance operations. Laboratory tests were performed using a set of 16 parallel limestone slabs 1.2 mm thick, packed within a horizontal PVC filter (2 m in length and with a cross-section of about 0.2 m × 0.2 m) that reproduced the confined groundwater flow at the Nardo` site (24). The secondary effluent from Bari West wastewater treatment plant was passed through the limestone filters, and the outflow was recycled back into the inlet section (at 1.2 L/h) to obtain the flow conditions (i.e., velocity, temperature, and pH) in the fractures similar to those found at the Nardo` site (90-91
FIGURE 3. Normalized counts given by the analytical solution (5) and corresponding somatic coliphage (solid/open circles and solid squares) breakthrough curves derived from measurements in water samples during pathogen transport in horizontal fractures from the sinkhole to well SP4 (500 m) in the 1st and 2nd tests at the Nardo` site and in the IRSA horizontal PVC filter. m/d; 15-16 °C; pH 7.4-7.7). The variation in the concentration of somatic coliphages over time (C/Cmax) in the outflow water from the filter and in Nardo` well SP4 were determined. During simulations F(C*) was assumed equal to 1 in every fracture, as the fractures were initially free of virus particles, for the specific initial conditions selected during tests. The model results given by eq 5 have been plotted in Figure 3 together with coliphage counts detected in groundwater, for the applied operational conditions. The best fit solutions derived from the proposed model were very sensitive to changes in rr and rf. These coefficients were therefore considered to be the model parameters whose adjustment could best be used for the proper calibration of the analytical solution (5) at the Nardo` site. Model solutions also provided inactivation rates λ* (0.075 d-1) and λ (0.015 d-1), deposition coefficients (κ) 32 × 10-8 m (at 16.6 °C) and 16 × 10-8 m (at 15 °C) at Nardo` and 1.1 × 10-8 m in the laboratory (at 17.3 °C), and virus detachment rate coefficients (see Figure 3). The model outputs yielded phage curves that approximated the experimental measurements rather better than the traditional approaches that use a constant inactivation rate (first-order exponential model). Even though seven parameters (λ, λ*, rr, κ, F(C*), b, and jτ) may limit the application of the proposed model with respect to traditional approaches, the rate obtained using a first order phage reduction can be very different (50% on average) from the average rate obtained using eq 8. The average rate of phage reduction in the field was reduced considerably (i.e., by half) by decreasing the groundwater temperature from 16.6 (2nd test) to 15 °C (1st test). In contrast, the average virus reduction rate obtained in the laboratory was lower than those determined in the field, even though the water temperature was higher (17.3 °C). This confirms that the transport of viruses, and their attachment on fracture surfaces, are not independent processes and may be affected significantly by water chemistry and velocity, hydraulic conductivity (8), geochemical heterogeneities (22), and virus collision efficiency. Furthermore, the rates of virus reduction are also dominated mainly by the history of the deposition processes over a long time period (i.e., clogging or hydraulic conductivity reduction) (25), which increased the efficacy of the natural filtration of groundwater at the Nardo` recharge site. Consequently, although the PVC filter employed in the present study was devised to simulate the conditions at the VOL. 45, NO. 2, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
639
FIGURE 4. (a) Virus reduction rates during wastewater transport in fractures and (b) virus breakthrough curves (solid squares and open circles) retrieved in the well SP3, 320 m downstream of the injection site. Best fit of model results using both constant (straight lines) and a time-dependent (dotted and solid lines) virus reduction rates. Nardo` fractured aquifer as closely as possible, there are differences between the laboratory and field conditions. The injection at that Nardo` aquifer started 18 years previously, causing biomass accumulation onto fracture surfaces that affected the recent filtration experiments. The values of κ, rr, λ, and λ* determined by field tests have also been proposed for use in further investigations of virus reduction at the Nardo` site.
Results and Discussion Virus Modeling. Solutions of the Lagrangian particle tracking model (24) were used to determine the elapsed time jτ required
by the majority of numerical particles to reach the observation well SP3 (1.1 d). Figure 4a-4b show virus rate reduction and predicted counts (C/Cmax) vs time t ) τ-τj. The model outputs successfully fitted the observed breakthrough curves of somatic coliphages in the sampled water at well SP3 during Nov-Dec 2006 and May-June 2007 by using the appropriate F(C*) ()0.1) value. The model was very sensitive to the rf changes from 4.64 × 10-8 m2 d-1 (at 15 °C) to 9.31 × 10-8 m2 d-1 (at 16.6 °C) (see Figure 4a). It should be noted that when the virus reduction decreases according to a first-order rate (0.14 or 0.28 d-1), the ratio C/Cmax is described by a straightline in Figure 4b, which deviates from experimental data. Figure 4a shows that a 50% variation in rf yields changes up to three times in the virus reduction rate. The range of variability of the time-dependent virus reduction rate decreases with decreasing groundwater temperature. Microbiological Monitoring. Tables 1, 2, and 3 show the average concentration of pathogen indicators and main pathogens (protozoa and viruses) in both injectant and groundwater at the Nardo` site (see Table SI-1 in SI, for the chemical constituents of the groundwater). The quantitative results of the analyses of the viruses shown in Table 3 refer to the sampling of February 2007, when the maximum number of positive samples was observed. A 99% (or 2 Log disinfection credits) removal of Giardia cysts from the SP2 to SP4 groundwater stream (Table 2) was observed, on average, and all the groundwater samples tested negative for HAV. The concentration of Cryptosporidium oocysts decreased from SP2 to SP3 by 36% on average, although a high concentration (0.77 oocysts/L) was found at SP3 during sampling of May 2007. A removal higher than 94.8% of Cryptosporidium oocysts, on average, were also tested from the SP2 to SP4. HAV has never been detected in the Salento groundwater, whereas Enteroviruses (Table 3) were tested positive in only four (of the total of 37) samples using PCR. Noroviruses were the most frequently retrieved viruses in the water samples analyzed, even if they were only retrieved in the winter samples. This seasonal occurrence is confirmed by other data described in the literature (26). In fact, these viruses are one of the major causes of nonbacterial gastroenteritis (winter vomiting disease), which is common during cold seasons in developed countries. Of the 37 Nardo` samples analyzed, only eight (22%) tested positive for enteric
TABLE 2. Giardia Cysts and Cryptosporidium Oocysts Retrieved at the Nardo` Site during 2006-2007 in Variable Size Volumes (from 10 to 30 L) of Sampled Water sampling points Cryptosporidiumoocysts (oocysts/L)
Giardiacysts (cysts/L)
SP1-WWTP (effluent) SP2-Sinkhole SP3-GW 320 m SP4-GW 500 m SP1-WWTP (effluent) SP2-Sinkhole SP3-GW 320 m SP4-GW 500 m
Jan-2006 0.15
