APPI.IKD AND ENVIKONMI-.NTAI. MH KOHIOIIK.V, May 1W2. p. INN- IMh
IX)W.2240/V2;imfiU«M)! Y. OUYANGt
          Department of Soil and Environmental Sciences; University of California, Riverside. California V252I

                                  Received V September IWl/Accepletl 21) February 1W2

          As • mull of the recently proposed mandatory groundwater disinfection requirements to Inactivate viruses
        in potable water supplies, there has been Increasing interest in virus fate and transport in the subsurface.
        Several models have been developed to predict the fate of viruses In groundwater, but few include transport in
        the unsaturated lone and all require a constant virus inactlvation rate. These are serious limitations in the
        models, as It has been well documented that considerable virus removal occurs in the unsaturaled zone and thai
        the inactivatkra rate of viruses Is dependent on environmental conditions. The purpose of this research was to
        develop a predictive model of virus fate and transport in unsaturaled soils thai allows the virus inactlvation rale
        to vary on the basis of changes In soil temperature. The model was developed on the basis of the law of mass
        conservation of a contaminant in porous media and couples the Hows of water, viruses, and heal through the
        soil. Model predictions were compared with measured data of virus transport in laboratory column studies
        and, with the exception of one point, were within the 95% confidence limits of the measured concentrations.
        The model should be a useful tool for anyone wishing to estimate the number of viruses entering groundwater
        after traveling through the soil from a contamination source. In addition, model simulations were performed
        to Identify parameters that have a large effect on the results. This Information can be used to help design
        experiments so that important variables are measured accurately. t
  The significance of viruses as agents of groundwatcrbornc
 disease in the United Stales has been well documented (3,4).
 The increasing interest in preventing groundwater contami-
 nation by viruses and other disease-causing microorganisms
 has led to new  U.S. Environmental  Protection Agency
 regulations regarding groundwaicr disinfection (21). the de-
 velopment of wellhead protection zones, and stricter stan-
 dards for the microbiological quality of municipal sludge (20)
 and treated effluent (2) that arc applied to land. For many of
 the new regulations, a predictive model of virus (or bacterial)
 transport would be  helpful in  the implementation process.
 For example,  such a model could be  used to determine
 where septic tanks should be placed or  where land applica-
 tion of sludge  or effluent should be practiced  relative to
 drinking water wells to minimize negative impacts on the
 groundwater quality. Another application of microbial trans-
 port models is  related to the groundwater disinfection rule
 (21). Water utilities wishing to avoid groundwater disinfec-
 tion may use a pathogen transport model to demonstrate that
 adequate removal of viruses  in the source water occurs
 during transport to the wellhead.
  Several models of microbial transport  have been devel-
 oped during the past 15 to 20 years (6. 7,  11.  12.17, 18. 23.
 27). The models range from the very simple, requiring few
 input parameters, to the very complex,  requiring numerous
 input parameter*. For many of the more complex models (7.
 11. 23), the data required for input arc noi available except
 for very  limited  environmental conditions. They may be
 useful for research  purposes but would be  impractical  for
 widespread use. The potential  applications of these models
 also range considerably, from being useful only for screening
 purposed on a regional scale (27) to predicting virus behavior
 at one specific location (6,13,  18). One  limitation of almost
 all  of  these models is that  they  have  been developed to
  ' Corresponding author.
  t Present iddrcss: Soil Science Department, University of Flor-
 ida. Gainesville. FL.
describe virus transport in saturated soils (i.e., groundwa-
ter). However, it has been demonstrated many times thai the
potential for virus removal is greater in the unsaturated zone
than in the groundwater (°, 10,  14).  If the viruses arc
transported through the unsalurated zone before entering the
groundwater, then  neglecting the unsuturated  /one and
assuming that the viruses immediately  enter the  saturated
zone in a model of virus transport could  lead to inaccurately
high predictions of virus concentrations at the site of inter-
est. This omission would be especially  significant in areas
with thick unsaturated zones, such as those in many western
states.  The one transport model (IK) that  h;is reportedly been
developed for predicting virus transport in variably saturated
media  is not  specific  for viruses but can be used for any
contaminant.  In addition, it has not been tested with data of
virus transport in unsaturated soil.
  Another, more important limitation of published models of
virus transport is that none of them has been validated by
using actual data of virus  transport  in unsaturated soils.
Most models  arc developed on the basis of theory and are
fitted to data obtained from one or two experiments. Rarely
are they tested by applying the model to data collected under
a variety of conditions and by then determining how well the
model  predicts what has been observed  in the laboratory or
field without any fitting or calibration of the model.
  The  purpose of this research was to develop a model that
can be  used to predict virus movement from a contamination
source  through unsaturated soil  to the groundwater. The
model  was tested by comparing the model predictions with
the results of laboratory studies. Several model simulations
were then performed to determine the  effects of different
input parameters on model predictions.


             MATERIALS AND METHODS

  Model development. The computer model, VIRTUS (virus
transport in unsaturated soils), is a one-dimensional numer-
ical finite difference code written in FORTRAN  program-
ming language. It simultaneously solves equations dcscrib-
                                                       IMN

-------
1MO
YATES AND OUYANCJ
                               Am.. ENVIRON. Mit KOHIOI..
ing the flow of water, viruses, unit heal through unsaturatcil
soil under different climatic conditions. The equation used to
calculate the transport of water through the soil is
                        & (TV! (e -
                                    - »>K'l           (l)
where / is time (in hours), p» is the density of water (in grams
per cubic centimeter), 6 is the volumetric soil water content
(in  cubic centimeters per cubic centimeter), iCJ(T) is the
density of water vapor at saturation al T (in grams per cubic
centimeter), T is temperature  (°C). h is the relative humidity
at (he atmosphere-soil interface (dimensionless), r is the soil
porosity  (in cubic centimeters of soil voids per cubic centi-
meter of soil), V, is the velocity of water in the liquid phase
(in centimeters per hour), and Vy is the velocity of water in
the vapor phase (in centimeters per hour). Heat transport
through the soil is calculated  by using the equation
     .1

     ill
     -V-
                         /« -t-flH,, -Mr- »)//sv|
where <•„,,„, is the specific heat of the solid (in calories per
gram per degree Celsius) (1 cal = 4.184J), |>MllK, is the density
of the solid (in grams per cubic centimeter). CM, is the
specific heat of (he air (in calories per gram per degree
Celsius). |>.,,, is the density of the air  (in grams per cubic
rentimeler). CM is the specific heat of the water (in calories
l/cr gram per degree Celsius), //,„ is  the transfer of heat by
conduction through the soil particles (in calories per square
centimeter per hour), //,, is the transfer of heal by conduc-
tion and convection in the liquid-phase water (in calories per
square centimeter per hour), and //„ is the transfer of heal
by conduction in the vapor-phase water and by transport in
the form of latent heat (in calories per square centimeter per
hour). The  equation governing  the transport of  viruses
through the soil is given by:
  ill
           HC',)
                          /> /   i>Ci\
                         -   0/>— - J -
                         in. \    i>z J
                                  - e/Ti              (3)

where (>,, is the bulk density of the soil (in grams per cubic
centimeter), C\ is the concentration of viruses adsorbed to
the soil (in PFUs per gram of solid), C, is the concentration
of viruses  suspended in the liquid  phase  (in  PFUs  per
millililer), 1) is the hydrodynamic dispersion coefficient (in
square centimeters per hour). M., is the inactivation rale of
viruses in the liquid phase (per hour). m is the inactiyalion
rale of adsorbed viruses  (per hour)./is the filtration coeffi-
cient (per  centimeter), and ; is the  position in space (in
centimeters). The derivations of these equations arc given by
Ouyung (13) and Yalcs ct al. (29).
  The processes used in the model to describe virus fate and
transport include advection (transport by the bulk movement
of water), dispersion (spreading out  of the viruses as they
move  around soil particles), adsorption, inactivation. and
filtration. A complete discussion of these factors and their
effects on microbiul transport has been published recently
(2H). Some of the specific features of the model will now be
described.
                                                      In the model, adveciion and dispersion of the virus parti-
                                                   cles are allowed  to  vary  as the viruses are transported
                                                   through  the soil profile.  In other words, (he rate at which
                                                   viruses are transported through the soil varies on the basis of
                                                   the velocity of the water, which depends on the flow of heat
                                                   through the system, among other factors. Another attribute
                                                   of VIRTUS is that the user may input different virus inacti-
                                                   vation rates for viruses that are adsorbed to the soil particles
                                                   as compared with freely suspended viruses, if that informa-
                                                   tion is known.
                                                     One important feature of the model is that the inactivation
                                                   rate does not have to remain constant throughout the simu-
                                                   lation. Because the model simulates the flow of heat through
                                                   the soil, it  allows one to compute a  new value  for any
                                                   heat-dependent variable as the temperature changes in the
                                                   soil profile.  It has been well documented that virus inactiva-
                                                   tion rates arc temperature dependent (8, 16, 24). An equation
                                                   describing the relationship between virus inactivation rates
                                                   and subsurface temperatures has been developed previously
                                                   (25) and is
                                                                          u =  -0.181 + (0.0214  x 7)
                                                                                                        (4)
where |i is the inactivation rate of the viruses (in log,,, per
day) and  7 is the temperature (°C).  Thus, whenever the
temperature of the soil changes, VIRTUS calculates a new
virus inactivation rate on the basis of this equation. The user
may specify the virus inactivalion rate to he a constant or a
function of any of the variables in the program. Equation 4
was used in several of the examples that will be presented
herein.
  Model testing. The model was tested for its ability to
predict  virus movement measured in laboratory column
studies. Three data sets that contained sufficient information
about the soil properties for the  model were  obtained. In
examples 1 and 2, the data were obtained from virus trans-
port experiments using saturated soil columns conducted by
Grondin at  the  University  of Arizona, Tucson (5).  For
example 3,  the  data were obtained  from virus  transport
experiments using unsaturatcd soil  columns conducted by
Powelson at  the University  of Arizona,  Tucson, and  re-
ported by Powelson ct al. (14). The data used as model input
for  each example arc listed in Table 1.
  In each case,  the  model was run by using  input values
measured or reported by the respective investigator. Model
predictions were then compared with the virus concentra-
tions measured as a  function of soil depth and lime in the
laboratory.
  Model simulations. Several features of the  model were
demonstrated by using data for two different  soil types, a
loam (example 4) and a sand (example 5). Some of the input
data for these examples arc shown in Table 2. Soil data were
obtained from Ouyang (13) for the Indio loam and from Ungs
el al. (19) for the Rchovot sand. Virus data were obtained
from several sources (1, 6, 14, 26) reporting virus transport
characteristics in soils similar to those used in the model. In
all simulations, water was added to the soil columns at a rate
of 0.1 cm h"' for 6  h. The concentration of viruses in the
influent solution was 105 PFU ml"1.
  In example 4.  the effects of three  different virus inactivu-
tion rales on model predictions were determined. For exam-
ple  4a, the virus inactivation rate varied as a function of the
soil temperature throughout the simulation. Virus inactiva-
lion rates were calculated by using equation 4. For examples
4b  and  4c. the virus inactivation rales were calculated  for
constant soil temperatures of 10 and 2S°C. respectively. In
these three examples, virus inaciivation was calculated only

-------
VOL. 58. W2
A MODI-L OF VIRUS TRANSPORT IN  UNSATURATliD SOILS
                                                                                                              1M1
                                        TABLE 1. Data used for model testing*
Properly
Soil type
Soil hulk density (g cm ')
llydrodynamic dispersion (cnr h 'l
Soil water content (cm' cm ')
Average water velocity (cm h ')
Soil column length (cm)
Soil adsorption coefficient (A!., => CJC,\ (ml g of soil '.)
Virus type
Virus inactivation rate (login day ')
Filtration coefficient (cm )
Input virus concentration (PFU ml ')
Simulation lime

Cxjmplc 1
(inivelly sand
1.65 g
7X
0.26
48.3
1(1(1
-0.054
MS2 coliphage
0.082
0
6..1 x 10'
48 min
Input value
Example 2
Gravelly sand
1.65
5«
0.26
28.3
100
-0.07.1
MS2 coliphage
0.056
0
8.37 x 10'
48 min

lixample .1
Ld;iniy line sand
1.54
•J2.24
Variable with depth
1.54
100
0.27
MS2 coliphage
2.00
0
10'
4 days
  " Krom (iromlin (Jl and P>rwelM>n cl *l. (14).
 for the freely suspended viruses, while the inuctivutiun rule
 (if  viruses adsorbed to soil  particles, u,,, was zero.  In
 example 4d,  the  inuctivution rate of adsorbed viruses wus
 specified to be one-half of the rate for viruses suspended in
 the water, u.,, which changed as a function of soil tempera-
 ture (i.e., same as example 4a with a u,, of 0.5u,,).
   Example 5  simulates the  transport of viruses through a
 sandy soil. In this example, the virus inactivation rate for
. freely suspended viruses changed as a function of tempera-
 ture as described in equation 4 with a u., of 0.

                       RESULTS

   Examples 1 and 2.  Figure  1 shows the predicted  virus
 concentrations at several depths after 4K min of transport in
 u saturated column of gravelly sand.  The model predictions
 were close to the measured virus concentrations and  in all
           cases fell within the 95% confidence limits of the measured
           data. In the second example, the model predictions were
           within the 95% confidence limits of the measured data at all
           points except the  UN)-cm depth (Fig. 2). Compared to the
           measured virus concentrations, the model overpredictcd the
           concentration of viruses that would be present in the column
           outflow. Grondin  (5)  measured 0  PFU of viruses alter 4K
           min. while the model predicted, on the basis of Crondin's
           data, that the virus concentration would be 341 PFU ml  '.
              Example 3. Virus transport in an unsaturaled soil column
           of loamy line sand,  with measured  values provided  by
           Powelson el ul. (14),  is depicted in Fig. 3. The agreement
           between model predictions and the  observed dulu is very
           good in this case.  The  model  predicted that the  virus
           concentration in the column outflow after 4 days would he
           3.54 login PFU ml"', while the measured concentration was
           3.78 log,,, PFU ml  '.
                                       TABLE 2. Data used for model simulations
                                                                                      Inpul value"
riu|iviijr
Soil type
Soil hulk density (g cm ')
Hydrodynamic dispersion
Initial soil water content (cm1 cm ')
Residual soil water content (cm1 cm " ')
Initial soil temp (°C)
Saturated hydraulic conductivity (cm h ')
Soil column length (cm)
Soil adsorption coefficient (K,, - CVC'i) (ml g of soil' ')
Virus type
Virus inactivation rale (free) (|t,)
Example 4a
Example 4h
Example 4c
Example 4d
Virux inactivulion rale (adsorbed) 
-------
If 12     YATHS AND OUYANG
                                AI'F'I.. I'NVIKCIN. MlCKOHIOI .
 Depth (cm)
         0


        JO


        40


        «0


        M


        WO


        120
                    1        2        3
                        Concentration (tog phi/ml)
  FIG. 1.  Comparison of mixJcl predictions with experimental dalu
of Grondin |.S) for  a  saturated, gravelly sand  soil  (example 1).
Ninety-live percent  confidence limits were calculated from seven
replicates.
Depth (cm)
        0


       20


       40


       ao


       SO


     .  100


       120
                         4              6
                       Concentration (log plu/ml)
  FIG. 3. Comparison of model predictions with experimental dad
of Powclson el al. (14)  for a loamy fine sand soil  (example 3).
Ninety-five percent confidence limits were calculated from seven
replicates.
   Example 4. Virus concentrations in the UX)-cm-long col-
umn of loam soil predicted by.using a variable inactivation
rate arc shown  in Fig. 4. Four different curves arc shown,
representing snapshots of the virus concentration profile in
the column after 6. 24, 72, and 120 h of transport. Figures Sa
and b show the effects of the different inuctivation rates on
model predictions of virus  transport. In Fig. Sa, the differ-
ence  in the  concentration  of virus particles  predicted by
using a  variable inactivation rate  and the constant rate at
WC is shown. The difference between predicted concentra-
tions by using the variable,  temperature-dependent inactiva-
tion rate and the constant rate at 2S°C is shown in Fig. 5b.
   The differences in virus  concentrations predicted by the
model when the rule of inactivation of adsorbed viruses is
zero compared  to when the rate  of inactivation of adsorbed
viruses is assumed to be one-half that of the free viruses arc
shown in Fig. 6.
   Example 5. Model predictions of virus transport in  a soil
column  of Rchovot  sand with  the virus inactivation rate
 ocprn (om)
           0.1        2       3        4        (
                        Concentration (log plu/ml)
  FIG. 2. Comparison of model predictions with experimental data
of Grondin (5)  for a saturated, gravelly Hand Mill (example  2).
Ninety-five percent confidence limits were calculated from seven
replicates.
calculated  as a function of  temperature as described  in
equation 4 arc shown in Fig. 7.

                      DISCUSSION
  Model testing. The ultimate measure of the usefulness of a
model as a predictive tool is its ability to accurately predict
field observations  of virus transport  under a  variety  of
environmental conditions. However, most models that have
been developed to predict microbial transport have not been
tested by using lie Id or laboratory data.  There arc a few
exceptions to this. For example, Tcutsch ct al.  (17) devel-
oped a one-dimensional model to describe microbial trans-
port that includes decay, growth, filtration, and adsorption.
The model predictions compared closely with the measured
         Virus   Concentration   (PFU/ml)
         10°     10'      ioj      io3      10*      10*
  o

  4-t
  fr
  Q
                                                                   20
                                                                   40
                                                                   60
                                                                   80
       100
  FIG. 4. Virus concentration as a function of soil depth when u
temperature-dependent inactivation rale was used with  an Indio
loam noil (example 4n).

-------
Vol.. 58. 1W2                                   A MODEL OF VIRUS TRANSPORT IN UNSATURATED SOILS     1M.1


   a       CCT  -   C10 (PFU/ml)                            C^  -   C^  (PFU/ml)
        -12000    -9000    -6000   -3000
        0
    O
   ^  10
   Q

   •6   20
        30
 •—•t •  24  h
 «-«• I »  72  h
    ' t -  120  h
-  C25 (PFU/ml)
                  2000    4000     6000     8000
   •5   20
   CO
        30
                      •—•t  » 24  h
                      »-»t  - 72  h
                           I  - 120  h
  FIG. 5. (a) Differences in predicted virus concentration when a
temperature-dependent (Ccl) or a constant (C,,,) inactivation rate
was used with  an  Indio loam soil (example 4a versus 4b). (b)
Differences in predicted virus  concentration when a temperature-
dependent (C,.,) or constant (C,,) inactivation rate was used with an
Indio loam soil (example 4a versus 4c).
results of a high-Row-ratc experiment of MS2 transport.
However, at low flow rates, microbial behavior could not be
.simulated closely by  using the name transport equation.
Harvey and Garabcdian (7) simulated bacterial transport by
using a colloid filtration model that had been modified to
include advcction, storage, dispersion, and adsorption. They
compared model predictions with measurements of bacterial
transport in a sandy aquifer in Cape Cod. Mas.s. While the
model wan able to simulate the bacterial transport measured
at a sampling point at a depth of 9.1 m, model predictions for
a sampling point at a depth of 8.5 m were not very close to
the measured concentrations, especially at later limes.
                                                    1000   2000  3000   4000  5000
    t «  24  h
«-« t •  72  h
    t =  120  h
                                                           FIG. ft. Effects of assuming no inaclivalion of adsorbed viruses
                                                          (C'nu.) or of assuming a nonzero inaclivalion rale of adsorbed viruses
                                                          (£',„,) '>n model predictions for an  Indio loam soil (example 4a
                                                          versus 4d).
                                      Both of these models  were developed for use by  the
                                    investigators to simulate their own data. In the case of the
                                    colloid filtration model, extensive fitting of the required input
                                    parameters was performed by calibrating different solutions
                                    of the transport equation to the observed bacterial break-
                                    through curves (7). Thus, while these models may be able to
                                    simulate the data of the investigator reasonably well, they
                                    may not  be able to predict  the results of the transport
                                             Virus  Concentration   (PFU/ml)

                                            K>°      10'      102     103     10*     10*
                                      u
                                       O
                                      to
                                      FIO. 7. Virus concentration as a function of noil depth when a
                                    temperature-dependent Inactlvalion rate wax used with a Rehovoi
                                    rand soil (example 5).

-------
1M4     YATES AND OUYANG
                                                                                          Am.. ENVIKON. MiotoHioi..
experiments uf other investigators. If a model is to be used
for purposes other than research, such as community plan-
ning or making regulatory decisions,  it must be able to
predict microbial transport by using data obtained by anyone
under a wide range of environmental conditions.
  Tim and Mostaghimi (18) attempted to simulate the results
of a saturated-flow column transport experiment using po-
liovirus I conducted by Lance and Gcrba (10). They used a
conventional equation for describing solute transport, i.e.,
the  advcction-dispcrsion  equation, in their  studies. The
difficulty encountered by these investigators was that insuf-
ficient data were reported by  Lance and Gcrba (10) to fulfill
the input requirements of the  model. Therefore, they had to
estimate values for the virus adsorption coefficient, the virus
inactivation rate, the saturated hydraulic conductivity, the
hydrodynamic dispersion coefficient, the moisture content at
saturation, and the average porosity of the soil. The model
simulation of virus concentrations compared closely  to the
measured virus concentrations in the top 80 cm of the soil
column; however, because so many of the input values were
estimated, it is difficult to assess the accuracy of the model.
   In this research, a model to describe virus transport was
developed on the basis of the factors known to affect virus
Cute in  the subsurface. A survey of the  literature was
conducted  to  locate data sets in which the investigators
made measurements of not only virus properties but also soil
and hydraulic properties. Three  data sets were located and
used to lest VIRTUS. No fitting or calibration of the  model
was performed; the data and measurements as reported by
the respective investigators were used  as model input.
   When the predictions of VIRTUS were compared with the
results obtained by Grondin (5) by using a saturated gravelly
sand column, the model predictions were within  the 95%
confidence limits of the measured virus concentrations for
one  trial (Fig. 1). For the second trial, the model predicted
that more than 3(X) viruses ml"' would appear in the column
effluent after 48 min. although none were detected  in the
laboratory  study (Fig.  2). The discrepancy between the
model predictions and the laboratory measurements may be
due  to  the  reported  value for  the adsorption coefficient
(-O.S4 ml g of soil"'). This value was not measured by the
investigators by using a batch adsorption isotherm study;
rather,  the value was used as a fitting parameter  for their
data. In the model, a  negative value for the  adsorption
coefficient would have the effect of transporting the viruses
at a more rapid rate  through  the soil (on average)  than the
average velocity of the water and resulted in viruses being
present in  the column effluent. If, in reality, there was
adsorption of the viruses to the soil  particles,  this would
retard their  movement through the column and result in no
viruses  being detected in the  outflow.
   In the case in which VIRTUS wan tested by using the data
of Powclson ct al. (14), model predictions were very close to
the measured virus concentration profiles (Fig. 3). However,
this is only one example of a comparison to one laboratory
transport study in unsaturatcd soil by using a single  soil type
and a single virus type. More testing of the model in required
before it should  be used for any purposes other  than re-
search.
   Unfortunately, in these examples, the temperature-depen-
dent inuclivation  rate capabilities of the model could not be
tested.  This is due to the fact  that the experiments were
conducted  under  constant temperature  conditions in the
laboratory,  and. thus,  the virus inactivation rate remained
constant (theoretically) throughout the  course of the exper-
iment. To tckt the capacity of the model to calculate new
virus inactivation rates as a  function of the  changing soil
temperature, data from a laboratory study  in which (he
temperature is allowed to change (and is closely monitored)
or from a field study in which the temperature is monitored
will  be required. This will  allow  an assessment  of the
capability of the  model to accurately calculate heat flow
through the soil,  which affects water flow (and thus virus
transport) as well as the rate of virus inactivation during
transport.
  Model simulations. In addition to being predictive tools,
models are useful for demonstrating the effects of different
variables on model  results. Because it  is not feasible to
perform experiments on all possible combinations of viruses,
soil types, and environmental conditions  to determine their
transport behavior, models can serve as a useful alternative.
The value of input variables can be easily changed, and the
results on model outputs can be determined.  For example,
the model can be  run by using different values for tempera-
ture  while holding constant all other values.  By using this
technique, a quantitative measure of the influence of tem-
perature on model results can be obtained. If it is shown that
a given variable  has a considerable  effect on the mode!
predictions, this indicates that experiments should  be de-
signed in such  a  way  that the variable is  measured accu-
rately. Several  factors that affect the  transport and fate of
viruses in the unsaturatcd zone, and which thus affect model
predictions, were investigated by using model  simulations
and arc discussed below.
  (I)  Effects of temperature-dependent inactlvalion  rates.
Most models of contaminant  transport consider the move-
ment of water and the transport of the contaminant in (heir
development and  assume that the thermal conditions in the
soil remain constant. In reality, under field conditions, this is
not generally the case. Temperature fluctuations in soil can
be considerable throughout  the course  of a 24-h period,
especially near the  soil surface. Because the effects of
temperature on virus inactivation rates in the environment
can be quite significant, it seems logical  to use a  model of
contaminant transport that also models heat flow.
  The effects of allowing the virus inactivation rate to vary
as a function of soil temperature in comparison  with the
effects of holding it constant arc graphically shown in Fig. 5a
and b. In the case where the virus inactivation rate was held
constant at  0.033 log,,, day"1 (10°C), the model predicted
higher concentrations of viruses than would be  predicted if
the inactivation rate was allowed to vary as  a  function of
temperature (Fig. Sa). The opposite predictions were ob-
tained in the case of a constant inactivation  rate of 0.354
log,,, day"1  (25°C), as shown in Fig. 5b. When the inactiva-
tion  rate  was  considered  to be a constant at  2S°C  an
underprediction in the concentration of viruses resulted as
compared with that predicted when the inactivation rate was
considered to be temperature dependent.
  The reasons for these predictions become apparent upon
observation of the predicted change  in soil temperature that
occurs as applied water is infiltrated through the soil column.
Figure 8 shows the soil temperature as a function of time for
the model simulations discussed for example 4 above. At the
soil surface, over a 24-h period, the soil temperature (which
started at 8.7*0) decreased to 3°C at 6 h during the addition
of cold water and increased to 3S°C at 12 h because of the
effects of solar radiation. Similar patterns would be expected
at the  5- and 10-cm depths, although  the magnitude of the
variation would not  be as large.  In example  4b,  the virus
inactivation rate was held constant at a value that would be
expected for constant 10°C soil conditions. The fact that the

-------
VOL. 58, 1992
                 A MODEL OF VIRUS TRANSPORT IN UNSATURATED SOILS
                                                                                1615
      40
  O  30
   0)
   a
       10
                           •—• 0  cm
12
24
                                         36
48
                        Time  (h)
  FIG. 8.  Soil temperature as a°'function of time for an Indio kxim
soil (example 4).
soil temperature rose above 10°C for more than 12 h in a 24-h
period resulted in a prediction of virus inactivation at rela-
tively high rates (compared to the rate at a constant temper-
ature of  10°C)  for that period. Overall, maintaining the
inactivation  rate  at a  constant value  had  the  effect of
increasing the predicted concentration of viruses that were
transported in the soil column  by more  than 4 orders of
magnitude (Fig. 5a).
   In example 4c, the soil temperature was considered to be
constant at 25°C; consequently, the virus inactivaiion was
maintained at a relatively high rale throughout the transport
process. In actuality, the soil temperature was at or above
25°C  for a relatively short period of time (less than 6 h), so
viruses were inactivated at or above that high rate for only 6
h in the  simulation where the rate was temperature depen-
dent. In this case (Fig. 5b), an assumption of a constant
inactivation rate would lead to a prediction that thousands of
viruses fewer than the actual  number (assuming that the
variable  inactivation  rate  simulation  predicts the  actual
number) would be transported in the column.
   The sensitivity of model predictions to changes  in the
temperature-dependent, inactivation rate was determined by
changing the inactivation rate while keeping all other varia-
bles constant. This sensitivity analysis showed that changing
the value of the inactivation rate by 50% resulted in a 33%
change  in  the  predicted concentration of  viruses being
transported through the soil. A high sensitivity of model
predictions to the virus  inactivation rate  has also been
observed by Tim and Mostaghimi (18) and Park et al. (12).
These results demonstrate the need to accurately monitor
virus inactivation and/or temperature during experiments of
virus transport in the subsurface.
   (U) Effects of inactivation rates for adsorbed vcnos those of
(rcc viruses. There have been reports in the literature of
differences in the measured rates of virus inactivation for
viruses that arc adsorbed to soil particles as compared with
those for viruses that  are freely suspended in the liquid
medium  (8, IS, 22). Therefore, this model was developed to
allow the user to input different values for inactivation rates
for viruses  in  these  two  states.  When  u value fur the
inactivation  rate of adsorbed viruses is specified. Ihc  model
calculates the number of viruses adsorbed at a given time on
the basis of  the adsorption coefficient specified by ihe user
and determines accordingly the number of viruses  inacti-
vated.
  It is difficult to obtain a quantitative value for the relative
difference between inactivation rates  for adsorbed viruses
and those for freely suspended viruses. For the purposes of
illustration,  a simulation with a value for adsorbed viruses
equal to one-half that of free viruses (temperature  depen-
dent) was compared with a simulation in which the inactiva-
tion  rate  for adsorbed  viruses was zero. As one  would
expect,  the concentration of viruses transported through the
soil column is larger when the solid-phase inactivaiion rate is
zero than when it is one-half the liquid-phase rate. The
difference increases with  time, as shown  in  Fig. 6. In a
system in which the inactivation rate of adsorbed viruses is
equal to that of free viruses, the differences would he even
greater.
  This example demonstrates the importance of knowing the
inactivation  rate  for viruses in the  adsorbed and  liquid
phases.  If the inactivation rate  for  adsorbed viruses is
actually lower than that of suspended viruses, it would be
important to incorporate that information in a model so that
accurate predictions could be  made.of virus  concentration
profiles. If the model assumes the same inactivation rate for
all viruses,  it would predict that  fewer viruses arc being
transported  than Ihc actual number.
  (til) Effects of soil type. A simulation of virus transport by
using data for a Rchovot sand was run to illustrate the effects
of soil properties on transport. The Rchovot sand has a much
higher hydraulic conductivity (Table 2) than that of the Indio
loam, and, thus, water and contaminants can  move through
this soil more rapidly. As shown in Fig. 7, the viruses were
transported  more rapidly and in higher concentrations in this
soil than in Ihe loam soil of ihc previous examples. After 6 h,
the viruses in the loam soil had been transported only 11 cm
(Fig. 4), in comparison to more than 35 cm in the sandy soil
(Fig.  7). The differences between the two columns become
more apparent at longer times: after 5 days, approximately
30 viruses ml~' had been transported 15 cm in the loam soil,
whereas more than 10r viruses ml"' were being recovered in
the sand column effluent after the same length of time.
  Another reason for the relatively higher concentrations of
viruses being transported through this soil, in addition to the
higher hydraulic conductivity, is related to the adsorption
coefficient. For this sand, on the basis of reported values for
virus adsorption to other sandy soils, an adsorption  coeffi-
cient of zero was chosen. Thus, the rate at which the viruses
were transported through the  soil was  not decreased as a
result of adsorption to the soil particles, unlike the case for
the loam soil.
  Condusioas.  A model of virus transport, VIRTUS, that
simultaneously solves equations describing the transport of
water, heat, and viruses through the unsaturatcd zone of the
soil  has been  developed.  The effects of a  temperature-
dependent inactivation rate versus a  constant inactivation
rate were shown to be considerable in terms of the concen-
trations of viruses that are predicted to be transported, in
addition,  it  was shown that different  inactivation rates  for
adsorbed versus freely suspended  viruses may have  a con-
siderable effect  on model  predictions. More data  on the
relative inactivation rates for viruses in these  two states ;ire-
necessary so that model  input "ilucs  arc as accurate as
possible.

-------
1616    YATES AND OUYANG
                                 AHPL. ENVIRON. MICROHIOI..
  VIRTUS was tested by using three data  sets  obtained
during  laboratory  studies of coliphagc transport  and was
found to produce reasonable predictions in comparison with
measured  results.  However, before this or any model of
contaminant transport can be used with confidence for any
purpose other  than  research,  considerable  testing is re-
quired. VIRTUS must be tested by using field data collected
in a wide  variety of environmental and hydrogeologic set-
tings, so that its limitations can be assessed. Few, if any,
data  sets  containing both virus data  and  the appropriate
hydrogeologic data are currently available so that this, or
any,  model can be tested. More transport  studies using
human viruses  that  have been implicated in waterborne
disease outbreaks and bacteriophages must be conducted to
assess the appropriateness of using phages or other micro-
organisms as surrogates for animal viruses in environmental
fate studies.

                   ACKNOWLEDGMENT

  This  research was  supported by intcragency  agreement DW
12933820-0 from the R. S. Kerr Environmental Research Labora-
tory,  U.S.  Environmental Protection Agency.

                       REFERENCES
  1. Biles,  R. C., C. P. Gerba. G. H. Grondln, and S. L. Jtnsea.
    1989. Bacteriophage transport in sandy soil  and fractured tun*.
    Appl. Environ. Microbiol. 55:2061-2067.
  2. California Department of Health Services. 1990.  Initial statement
    of reasons for proposed changes in  the regulations of the
    Department  of Health  Services pertaining to the use of re-
    claimed water other than for groundwater recharge and pertain-
    ing to  use of household'gray water at residences and text of
    proposed regulations. California Department of Health  Serv-
    ices, Sacramento.
  3. Craun, G. F. 1986. Waterborne diseases in  the United States.
    CRC Press, Boca Raton, Fla.
  4. Craun, G. F. 1990. Review of the causes of waterborne disease
    outbreaks, p. 1-22. In G. F. Craun  (ed.). Methods for the
    investigation and prevention of waterborne disease outbreaks.
    U.S. Environmental Protection Agency publication no.  EPA/
    600/1-90/QOSa. U.S. Environmental Protection Agency, Wash-
    ington, D.C.
  5. Grondtn, G. H. 1987. Transport of MS-2 and f2 bacteriophage
    through saturated Tanque Verde wash soil. M.S. thesis. Uni-
    versity of Arizona, Tucson.
  6. Grosser, P. W. 1984. A one-dimensional mathematical model of
    virus transport, p.  105-107. In Proceedings 'of the Second
    International Conference on Groundwater Quality  Research,
    Tulsa,  Okla.
  7. Harvey, R. W., and S. P. Garabedlan. 1991. Use  of colloid
    filtration theory in modeling movement of bacteria  through a
    contaminated sandy aquifer. Environ. Sci. Techno). 25:178-185.
  8. Hunt,  C. J., C.  P. GcrtM, and I.  Cech. 1980. Effects of
    environmental variables and soil characteristics on  virus sur-
    vival in soil. Appl. Environ. Microbiol. 40:1067-1079.
  9. Keswick, B. H., and C. P. Gerba. 1980.  Viruses in groundwater.
    Environ. Sci. Technol. 14:1290-1297.
10. Laoce.  J. C., and C. P. Gerba. 1984.  Virus movement in soil
    during  saturated and unsaturated flow. Appl. Environ. Micro-
   biol. 47:335-337.
11. Matlhess, G., A. Pekdegcr, and J. Schroeter. 1988. Persistence
   and transport of bacteria and viruses in groundwater—a con-
   ceptual evaluation. J. Contam. Hydro). 2:171-188.
12. Park,  N., T.  N. Blandford,  M. Y.  Corapcioglu,  and P. S.
   Huyakorn. 1990. VIRALT: a modular semi-analytical and nu-
   merical model for simulating viral transport in ground water.
   Office  of Drinking Water,  U.S.  Environmental  Protection
   Agency. Washington. D.C.
13. Ooyang, Y. 1990. Dynamic mathematical model of oxygen and
   carbon dioxide exchange between soil and atmosphere. Ph.D.
   dissertation. Oregon State University, Corvallis.
14. Powcbon, D. K., J. R. Simpson, and C. P. Gerba. 1990. Virus
   transport and survival in saturated and unsaturated flow through
   soil columns. J. Environ. Qua!. 19:396-401.
15. Sobsey, M. D., C. H. Dean, M. E. Knuckles, and R. A. Wagner.
   1980. Interactions and survival of enteric viruses in soil materi-
   als. Appl. Environ. Microbiol. 40:92-101.
16. Sobsey, M. D., P. A. Shields,  F. H. Hauchman,  R. L. Hazard.
   and L. W. Calon 111. 1986. Survival and transport of hepatitis A
   virus in soils, groundwater and wastewater. Water Sci. Technol.
   18:97-10^.
17. Ttntscb, G., K. Herbold-Paschkc, D. Touglanidou, T. Hahn, and
   K. Bolzenhart. 1991. Transport of microorganisms in the under-
   ground—processes, experiments, and simulation models. Water
   Sci. Technol. 24:309-314.
18. Tim, V. S., and S. Mostaghlml. 1991. Model for predicting virus
   movement through soils. Ground Water 29:251-259.
19. Vugs, M. J., L. Bocnma, and S. Akratanakul. 1985. OR-Naiuve:
   the numerical analysis of transport of water and solutes through
   soil and plants, vol. 4. Examples. Special report 754. Agricul-
   tural Experiment Station, Oregon State University, Corvallis.
20. U.S. Environmental Protection Agency. 1989. Standards for the
   disposal of sewage sludge; proposed rule. Fed. Regist. $4:5746-
   5902.
21. U.S. Environmental Protection Agency. 1991. Possible require-
   ments of the ground-water disinfection rule. 20 June 1991. Office
   of Ground Water and Drinking Water, Washington, D.C.
22. Vaughn, J. M.,  and E.  F. Landry. 1983. Viruses in soils and
   groundwater, p. 163-210. In G. Berg (ed.), Viral pollution of the
   environment. CRC Press, Boca Raton, Fla.
23. VUkcr, V. U, aid W. D. Bnrje. 1980. Adsorption mass transfer
   model for virus transport in soils. Water Res. 14:783-790.
24. Yates, M. V., C. P. Gerba,  and L. M. Kelley. 1985. Virus
   persistence in groundwater. Appl. Environ. Microbiol. 49:778-
   781.
25. Yates, M. V., and S. R. Yates. 1987. A comparison of geostatis-
   tical methods for estimating vims inactivation rates in ground
   water. Water Res. 21:1119-1125.
26. Yates, M. V., and S. R. Yates. 1988. Modeling microbial fate in
   the subsurface environment. Crit. Rev. Environ. Control  17:
   307-344.
27. Yates, M. V., and S.  R. Yates. 1989.  Septic  tank setback
   distances: a way to minimize virus contamination of drinking
   water. Ground Water 27:202-208.
28. Yates, M.  V., and S.  R.  Yates.  1990. Modeling  microbial
   transport in soil and groundwater. ASM News 56:324-327.
29. Yates, M. V., S. R. Yates, and Y. Ouyang. 1991. Developmem of
   a model of virus transport in unsaturated soils. Final report to
   R. S. Kerr Environmental Research Laboratory, U.S. Environ-
   mental Protection Agency, Ada, Okla.

-------