Journal of Contaminant Hydrology 53 (2001) 387 406

www.elsevier.com/locate/jconhyd

Progression of natural attenuation processes

at a crude oil spill site:
II. Controls on spatial distribution of
microbial populations
Barbara A. Bekins a,*, Isabelle M. Cozzarelli b, E. Michael Godsy a,
Ean Warren a, Hedeff I. Essaid a, Mary Ellen Tuccillo c
a
US Geological Survey, MS 496, 345 Middlefield Road, Menlo Park, CA 94025, USA
US Geological Survey, 431 National Center, 12201 Sunrise Valley Drive, Reston, VA 20192, USA
c
Department of Environmental Sciences, Clark Hall, University of Virginia, Charlottesville, VA 22903, USA
b

Received 7 March 2000; received in revised form 25 January 2001; accepted 13 February 2001

Abstract
A multidisciplinary study of a crude-oil contaminated aquifer shows that the distribution of
microbial physiologic types is strongly controlled by the aquifer properties and crude oil location.
The microbial populations of four physiologic types were analyzed together with permeability, porewater chemistry, nonaqueous oil content, and extractable sediment iron. Microbial data from three
vertical profiles through the anaerobic portion of the contaminated aquifer clearly show areas that
have progressed from iron-reduction to methanogenesis. These locations contain lower numbers of
iron reducers, and increased numbers of fermenters with detectable methanogens. Methanogenic
conditions exist both in the area contaminated by nonaqueous oil and also below the oil where high
hydrocarbon concentrations correspond to local increases in aquifer permeability. The results
indicate that high contaminant flux either from local dissolution or by advective transport plays a key
role in determining which areas first become methanogenic. Other factors besides flux that are
important include the sediment Fe(II) content and proximity to the water table. In locations near a
seasonally oscillating water table, methanogenic conditions exist only below the lowest typical water
table elevation. During 20 years since the oil spill occurred, a laterally continuous methanogenic
zone has developed along a narrow horizon extending from the source area to 50 60 m
downgradient. A companion paper [J. Contam. Hydrol. 53, 369 386] documents how the growth of
the methanogenic zone results in expansion of the aquifer volume contaminated with the highest

Corresponding author. Fax: +1-650-329-4463.

E-mail address: babekins@usgs.gov (B.A. Bekins).

0169-7722/01/$ - see front matter D 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 9 - 7 7 2 2 ( 0 1 ) 0 0 1 7 5 - 9

388

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

concentrations of benzene, toluene, ethylbenzene, and xylenes. D 2001 Elsevier Science B.V. All
rights reserved.
Keywords: Anaerobic environment; Microorganisms; Contaminant plumes; Ground water; BTEX; Biodegradation

1. Introduction
The establishment of aquifer microbial populations that degrade organic ground-water
contaminants is a key aspect to the success of natural attenuation of these compounds.
Studies of the degradative capabilities of aquifer microorganisms have documented spatial
variability in both degradation rates (Adrian et al., 1994; Nielsen and Christensen, 1994)
and ability to degrade specific compounds (Anderson et al., 1998). Such variability is
presumably due a number of underlying causal factors. These include variations in the
physical and chemical properties of aquifer materials, and in the aqueous concentrations of
contaminants and other important reactants (NRC, 1993). Significant progress in addressing these issues has been made by focusing on each aspect separately. For example, Webb
and Anderson (1996) used modeling to show how variations in aquifer physical properties
causes preferential contaminant migration along connected high permeability pathways. In
contaminated aquifers where the geochemical conditions have been exceptionally well
characterized, it has been noted that contaminant concentrations and associated geochemical conditions change over sub-meter scales (Cozzarelli et al., 1999). Moreover,
studies of microbial populations attached to the sediments in contaminant plumes indicate
that microbial physiologic types also vary on sub-meter scales (Bekins et al., 1999; Smith
et al., 1991).
Although it is clear that geochemical conditions and microbial physiologic zones can
vary on small spatial scales, studies are needed that relate this variation to the subsurface
chemical and physical properties. The results of these studies provide insight into how
spatial and temporal changes in aquifer conditions leads to a succession of microbial
physiologic types. This is important because the succession of microbial physiologic types
is associated with a change in the degradative capabilities (e.g. Krumholz et al., 1996). In
some cases, microorganisms with specific degradative capabilities are found in favorable
habitats that exist in only a portion of a plume (e.g. Anderson et al., 1998). Methods that
provide a basis for identifying specific habitats allow the capabilities of the associated
microbial populations to be characterized. An increased understanding of how redox zones
and associated microbial capabilities change over time facilitate forecasts of the long-term
fate of contaminant plumes.
This study is part of a larger effort to investigate the long-term fate of a plume of crudeoil contaminants. The goal of this work was to examine the processes that control the
spatial distribution of microbial physiologic types. Using the most probable number
(MPN) method, the microbial population distributions were estimated for four physiologic
types: aerobes, iron reducers, heterotrophic fermenters, and methanogens. In the same
locations, pore-water chemistry, nonaqueous oil saturation, sediment iron content, and
sediment permeability were also determined. The combined data set presented in this study

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

389

provides insight into controls on the patterns of microbial ecological succession at the
plume scale. These results can be related to observations of the long-term evolution of the
plume described in a companion paper by Cozzarelli et al. (2001) to show how the
microbial population succession has affected the contaminant plume.

2. Site description
The study aquifer, located near Bemidji, MN (Fig. 1), is a surficial formation of pitted
and dissected glacial outwash sediments. The sediments of the unconfined aquifer consist
of moderate to poorly sorted sandy gravel, gravelly sand, and sand with thin interbeds of
fine sand and silt (Franzi, 1988). Dillard et al. (1997) examined the permeability (k, in
square meters) using grain-size analyses and the Krumbein and Monk (1942) relation. The

Fig. 1. Map of the Bemidji research site showing the oil pipeline, location of the 1979 pipeline break, nonaqueous
oil body know as the north pool, water table contours, and locations of sample wells.

390

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

overall mean and variance of the log(k) distribution were 11.28 and 0.86, respectively.
The distribution was bimodal with 11% low permeability silty material and the remainder
higher permeability coarse-grained material. The water table is 6 10 m below land surface
and the direction of ground-water flow is to the northeast (Hult, 1984). The average linear
velocity of the groundwater is about 0.05 m/day in the silt layers and 0.15 0.5 m/day in
the coarser-grained sediment (Bennett et al., 1993).
The aquifer was contaminated with crude oil when a pipeline ruptured in August 1979.
After clean-up operations by the pipeline company, the unrecovered spilled oil infiltrated
into the aquifer. The oil is entrapped as a residual nonaqueous phase in the vadose zone
and also forms two bodies of oil floating on the water table (Dillard et al., 1997; Hult,
1984). The largest oil body (the north pool, Figs. 1 and 2) was estimated to contain
147,000 l of oil in 1998 (Herkelrath, 1999).
A set of papers (Baedecker et al., 1993; Bennett et al., 1993; Eganhouse et al., 1993)
described the geochemical evolution of the contaminant plume emanating from the north
pool. Starting in 1984, concentrations of both reduced iron and methane began to increase
in the anoxic zone immediately downgradient from the oil body. From 1986 to 1989,
profiles of dissolved organic compounds indicated the plume reached a quasi steady state
in which dissolution from the nonaqueous oil was approximately balanced by iron
reducing and methanogenic biotransformation (Baedecker et al., 1993; Lovley et al.,
1989). Fig. 2A and B contain contour plots of dissolved oxygen (DO) and dissolved
volatile organic carbon (VOC) measured in 1992 along the axis of the north pool plume
(A AV, Fig. 1) (Baedecker et al., 1993; Cozzarelli et al., 1996). Note that the anaerobic
core of the plume attains a minimum vertical thickness of about 1 m at the downgradient
edge of the nonaqueous oil body. Essaid et al. (1995) showed that the plume thickens
vertically below the oil body as groundwater flow is diverted around the area where
partial saturation with oil decreases the hydraulic conductivity. Observations on the
continued chemical evolution of the plume since 1992 are presented by Cozzarelli et al.
(2001).
On the basis of the pore-water chemistry, Baedecker et al. (1993) divided the aquifer in
the vicinity of the oil body into anoxic, transition, and background zones. Within the
anoxic zone, the primary anaerobic degradation processes are iron reduction and methanogenesis (Baedecker et al., 1993; Lovley et al., 1989). Bekins et al. (1999) used MPN data
from three vertical profiles and seven broadly spaced samples to map microbial
physiologic zones for aerobes, iron reducers, and methanogens in the aquifer near the
oil body (Fig. 3). Data from the three vertical profiles show that populations of aerobes,
iron reducers, fermenters, and total methanogens vary sharply over sub-meter scales. The
variations in microbial numbers are systematic indicating that different physiologic types
dominate in different portions of the contaminated aquifer.
Degradation of petroleum hydrocarbons under methanogenic conditions requires the
combined activity of fermenting bacteria that break down complex organics and two types
of methanogens (e.g. Schink, 1997). Locations in the Fig. 3 profiles where methanogens
increase together with fermenters indicate the presence of a methanogenic consortium
capable of utilizing petroleum hydrocarbons for growth. In the anoxic zone, there is also a
positive correlation between numbers of aerobes and methanogens. This implies that the
organisms from the anoxic zone that grew in our media under aerobic conditions are

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

391

Fig. 2. Cross-sections of 1992 concentration distributions in the contaminant plume below the north pool oil body
along the line A AV of Fig. 1. The area with greater than 10% oil saturation determined by Dillard et al. (1997) is
also shown. (A) Dissolved oxygen distribution modified from Cozzarelli et al. (1996) together with the microbial
sample locations. The double lines mark the locations of detailed vertical profiles A, B and C. (B) Total dissolved
volatile organic carbon are from Baedecker et al. (1996). Screened intervals and ID numbers of monitoring wells
are also shown.

capable of functioning both aerobically and anaerobically. In locations where these

aerobes increase with methanogens, the aerobes most likely function as fermenters in
a methanogenic consortium (Bekins et al., 1999). In the profiles, the numbers of bacteria
functioning as fermenters exceed those of methanogens by factors of 10 104. Such ratios
are comparable to those found in other studies of methanogenic consortia both in the
laboratory (e.g. Morgan et al., 1991) and in the field (Godsy et al., 1992). Thus, areas of
the aquifer classified as methanogenic contain culturable methanogens together with
increased numbers of fermenters or aerobes and decreased numbers of iron reducers.

392

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

Fig. 3. Vertical MPN profiles and interpreted cross-section showing the distribution of microbial physiologic
types in the anaerobic portion of the Bemidji plume. The vertical profiles correspond to Sites A, B and C in Fig. 2.
The interpreted cross-section is based on the vertical profile data combined with data for the point samples
marked on the plot (Bekins et al., 1999).

A correlation analysis of the combined MPN data set showed an inverse correlation
between iron reducers and methanogens indicating that methanogenic conditions are
present in areas where iron reduction is less favorable. Methanogenic zones have evolved
in two places: one within the area with separate-phase oil and a second below the oil in the

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

393

laterally migrating plume of contaminated ground water. In the laterally migrating plume,
there is a classic pattern of a methanogenic core progressing with depth to iron-reducing
and then aerobic conditions. Moreover, the establishment of methanogenic conditions
appears more advanced near the oil body at Sites A and B, than downgradient at Site C.
This is evident from the higher peak methanogen numbers and the lower minimal values
of iron reducers at Sites A and B compared to Site C. The anaerobic core of the plume
attains a minimum vertical thickness of about one meter at the downgradient edge of the
nonaqueous oil body (Site B, Fig. 2) (Cozzarelli et al., 1996). The microbial data from this
location indicate that the methanogenic zone centered at 422.5 m is especially narrow,
spanning a distance of only 25 cm, in contrast to the thicker 0.5 1 m methanogenic zones
at Sites A and C.

3. Methods
On the basis of the existing geochemical data, locations for three vertical profiles in the
anaerobic portion of the plume (Sites A, B and C; Fig. 2A) were chosen to characterize the
microbial population, pore-water chemistry, sediment iron content, and sediment permeability. In September of 1996, two cores with a horizontal separation of 1 m were collected
from Site B (cores 9609 and 9610). Each core was 2.3 m long, which was sufficient to
span the entire vertical extent of the anaerobic plume at this location. One core was
analyzed for the microbial population on the sediments, along with extractable iron [Fe(III)
and Fe(II)], and grain size. The second core was used for pore-water chemical analyses.
Because the drill rig depth measurements were imprecise, data from the two cores were
aligned vertically using the grain-size distributions. In August of 1997, five additional
cores were collected. Three cores were needed to span a vertical interval from the oil body
to the base of the contaminant plume at Site A (cores 9704 9706), and two cores were
sufficient to obtain a vertical profile of the plume at Site C (cores 9701 9702). In these
cores, the pore water was drained for chemical analyses and the sediments were analyzed
for microbial numbers, grain-size distributions, and extractable iron.
The cores were collected with a freezing drive shoe that freezes the bottom 10 cm of the
core in situ before retrieval thus preserving the position of the ground water with respect to
the cored sediments (Murphy and Herkelrath, 1996). The 47-mm diameter cores were
collected in clear polycarbonate liners pre-rinsed with methanol and de-ionized water. For
pore-water chemical analyses, water was drained from the core at 15-cm intervals. The pH
was measured in the field and the water samples were preserved for laboratory analyses of
hydrocarbons, dissolved organic carbon, methane, and major cations, according to the
methods described by Cozzarelli et al. (2001). The sediment grain-size analyses were
performed using the method described by Hess et al. (1992). Permeabilities were estimated
from sediment grain-size distributions using the method of Krumbein and Monk (1942).
The sediment iron was determined by 0.5 M HCl extraction as described by Cozzarelli et
al. (2001). This method was expected to extract the bioavailable Fe(III) as well as poorly
crystalline Fe(II) from the sediments.
Sediment samples were collected at 20 70 cm intervals and analyzed for populations
of aerobes, iron reducers, heterotrophic fermenters, and methanogens using the MPN

394

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

method. The methods of sediment sampling and MPN determination are described in
detail by Bekins et al. (1999). Briefly, approximately 10 g of sediment from the center
of the core was collected under a flow of oxygen-free nitrogen gas and added to a prereduced and anaerobically sterilized (PRAS) mineral salts solution containing a nonionic surfactant added to detach microbes from the sediment (Yoon and Rosson, 1990).
The tube was then sealed, mixed well, and allowed to stand for 2 h to allow penetration of surfactant into the sample. The tubes were then opened and sonicated to
dislodge the bacteria into the mineral salts under a flow of sterile oxygen-free nitrogen
gas.
Microbial concentrations were determined using a five-tube MPN analysis. Samples
were serially diluted by orders of magnitude into PRAS mineral salts solutions. Bacteria
capable of aerobic heterotrophic growth were enumerated using Standard Methods Broth
(BBL Microbiology Systems, Cockeysville, MD). Tubes with visible growth after
incubation at room temperature for one week were scored positive. Microorganisms
capable of anaerobic heterotrophic growth or fermentation were enumerated using PRAS
prepared Schaedlers Broth (Difco, Detroit, MI). Tubes with turbid growth or clumps of
particulates after incubation at room temperature for 1 week were scored positive.
Iron-reducing bacteria were enumerated using PRAS prepared media consisting of
sodium acetate and poorly crystalline iron in a mineral salts solution (modified from
Lovley and Phillips, 1986). After inoculation, the serum bottles were aseptically pressurized with a 70:30 mix of H2/CO2 to 140 kPa and then incubated for a minimum
of 6 weeks at room temperature. A serum bottle was scored as a positive if greater
than 2 mg/l of reduced iron (Fe2+) was present as determined with bipyridine. Microorganisms capable of methane production were enumerated on PRAS mineral salts
media. Acetoclastic and formate-utilizing organisms were enumerated under a nitrogen
and CO2 atmosphere with the addition of acetate or formate, respectively. Hydrogen
oxidizers were enumerated by aseptically pressurizing the serum bottles after inoculation
with a 70:30 mix of H2/CO2 to 140 kPa. The serum bottles were allowed to incubate for
a minimum of 6 weeks at room temperature. The presence of methane in the head space
was determined by Gas Chromatography/Flame Ionization Detection analysis (Godsy,
1980). Methanogenic numbers presented in this paper represent the sum of the numbers
obtained for the three types of methanogens at a given location. Although this will result
in double counting some methanogens, this method provides the best indication of the
maximum spatial extent of methanogenic activity. Results from this study of the
enumerations for the separate physiologic types of methanogens are published in Warren
et al. (1999).

4. Results and discussion

We now consider how the physical and chemical conditions in the aquifer have affected
the locations of the physiologic zones. Fig. 4 shows plots for Sites A, B and C of the ironreducer and methanogen MPNs together with the extractable sediment Fe(II) and Fe(III),
pore-water concentrations of ethylbenzene and the sum of m- and p-xylene, sediment
permeabilities, and pore-water Fe(II) concentrations.

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

395

Fig. 4. Vertical profiles for Sites A, B and C of microbial data for iron reducers and methanogens with interpreted
methanogenic zones of Fig. 3 shown as shaded gray bands. Plots of Fe(II) and Fe(III) concentrations from the
sediments; pore-water concentrations of ethylbenzene and m- and p-xylene; and permeabilities estimated from
sediment grain-size distributions and pore-water Fe(II) concentrations for each site are shown on the same vertical
scale. Uncontaminated background concentrations of sediment Fe(II) and Fe(III) are shown in the sediment iron
plots, averaging 6 and 25 mmol /g, respectively.

396

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

4.1. Extractable iron on sediments

In general, extractable concentrations of sediment Fe(III) are lower, and Fe(II) are
higher in the area of the aquifer affected by the plume than background values.
Sediment Fe(III) values at the base of the core at Sites B and C match background
concentrations indicating that the profiles span the entire anoxic portion of the plume.
The overall picture is consistent with the results of Tuccillo et al. (1999) who found
that microbial iron reduction of the petroleum hydrocarbons has resulted in depletion of
solid phase Fe(III) and precipitation of Fe(II) minerals on the sediments in the anoxic
zone. However, the microbial numbers and sediment iron concentrations together
provide a more comprehensive picture of the processes controlling the progression
from iron reduction to methanogenesis. First, in the methanogenic zones, extractable
iron and culturable iron reducers are still present. Moreover, detailed examination of
the sediment iron and microbial data shows that high concentrations of solid-phase
Fe(III) do not always correlate with high numbers of iron reducers and vice versa.
Below is a detailed discussion of these issues focusing on Sites A and B because these
sites have been anoxic the longest and have the greatest variability in sediment Fe
concentrations.
In areas with culturable methanogens there are also culturable iron reducers and
significant extractable Fe(III) suggesting ongoing low levels of iron reduction. The lowest
value of extractable Fe(III) found in all of the cores is 11 mmol /g which is almost 50% of
the average background concentration. Apparently, only about half of the Fe(III) extracted
on the aquifer sediments is used before methanogenesis begins. The sediment iron
extraction method recovers primarily ferrihydrite and only a small percentage of Fe(III)
from crystalline phases such as hematite, goethite and magnetite (Tuccillo et al., 1999). A
statistically significant inverse correlation between iron-reducers and methanogens
(Bekins et al., 1999), suggests that there is a continuum of decreasing iron-reducing
activity in proportion to the increase in methanogenic activity. This observation is
consistent with a range of iron phases in the aquifer that vary in their bioavailability.
Postma (1993) showed that simultaneous sulfate and iron reduction is thermodynamically
feasible if microorganisms are forced to use less soluble iron phases than the poorly
crystalline material that typically comprises hydroxypolymer coatings on mineral grains.
The Fig. 4 data suggest that a similar analysis is needed to determine whether simultaneous iron-reduction and methanogenesis is feasible under the geochemical conditions in
the contaminated portion of the Bemidji aquifer.
High numbers of iron reducers correspond to some of the lowest sediment Fe(III)
concentrations, at three locations within the oil body and just below it (Site A, 424 m and
Site B, 423.4 and 422.8 m). Preferential use of favorable Fe(III) phases, along with natural
variability of the Fe(III) phases on the sediments, may play a role. However, the proximity
of these locations to the separate-phase oil appears to be significant. There are several
mechanisms that may facilitate use of less-favorable iron phases near the separate-phase
oil. One possibility is that proximity to the oil provides access to more favorable substrates
such as toluene or o-xylene. Another important aspect is that organic ligands are present in
higher concentrations near the oil (Cozzarelli et al., 1994). Lovley et al. (1994)
demonstrated that organic ligands facilitate chelation of iron from solid Fe(III) phases.

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

397

Alternatively, the organic ligands may bind to aqueous Fe(II) reducing the effect of
product inhibition (Urritia et al., 1999).
Very low numbers of iron reducers correspond to moderate levels of Fe(III) in three
locations (Site A, 423.2 m and Site B, 422.4 and 422.9 m). These locations also have peak
values of sediment Fe(II) and active methanogens. The data suggest that high concentrations of sediment Fe(II) can inhibit iron reduction and provide an advantage to the
methanogens. Inhibition of iron reduction due to coating of Fe(III) oxides by precipitated
Fe(II) surface phases has been documented in the laboratory (Roden and Zachara, 1996;
Urritia et al., 1998). Cozzarelli et al. (1999) also noted at a gasoline-contaminated aquifer
that coating of iron oxide surfaces by reduced-iron phases apparently limited iron
reduction and allowed sulfate reduction to become more energetically favorable.
4.2. Flux of hydrocarbon contaminants
To illustrate the relationship between the flux of organic contaminants through the
aquifer and the microbial population distribution, the concentrations of ethylbenzene and
m- and p-xylene for the three sites are shown in Fig. 4. Ethylbenzene was selected because
it persists like benzene in the anoxic plume (Cozzarelli et al., 1990) but, unlike benzene, is
present at concentrations similar to m- and p-xylene. The behavior of m- and p-xylene is
intermediate compared to toluene, which degrades rapidly, and benzene, which degrades
more slowly (Eganhouse et al., 1993). Between Sites B and C, the average concentration
of ethylbenzene drops slightly from about 0.4 to 0.3 mg / l. In contrast, the concentration of
m- and p-xylene drops from 1 2 mg / l at Sites A and B to about 0.1 mg /l at Site C.
At Sites A and B, methanogenic conditions occur in two distinct zones in the anoxic
portion of the aquifer (gray shaded zones in Fig. 4). The upper zone corresponds to the
portion of the aquifer where separate-phase oil is present. The lower zone corresponds to
the center of the horizontally advecting plume below the oil. Both of these methanogenic
zones are associated with higher concentrations of ethylbenzene, m- and p-xylene, and
dissolved iron. The correspondence between methanogenic zones and higher concentrations of contaminants suggests that over time these areas have been subjected to greater
cumulative hydrocarbon fluxes.
In the zones where separate-phase oil is present at Site A (above 423 m) and Site B
(above 422.8 m), peak iron reducer numbers are lower than the maximum values found in
the aquifer and methanogens are present in all samples (Fig. 4). In this area, partial oil
saturations exceeding 10% reduce the hydraulic conductivity thus restricting downward
advection of hydrocarbons through the oil (Essaid et al., 1995). Eganhouse et al. (1996)
showed that the concentration of hydrocarbons in contact with the separate-phase oil is
controlled by equilibrium dissolution. Thus, although the advective flux of hydrocarboncontaminated water is low, the flux due to dissolution of the separate-phase oil is relatively
high in this zone.
Hydrocarbon flux in the plume below the oil occurs primarily by horizontal advection
of water contaminated by separate-phase oil located upgradient. At Sites A and B, a
methanogenic zone has developed in the center of the plume where the maximum
advective BTEX flux occurs. To illustrate this point, Fig. 5 shows the methanogen
numbers together with the computed flux of benzene, toluene, ethylbenzene and xylenes

398

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

Fig. 5. Computed flux of total BTEX plotted together with methanogenic and iron-reducing MPN data for the
three vertical profiles in Fig. 3. For profiles (A) and (B), only the intervals below the oil are plotted. Over these
intervals, there is a significant positive correlation between total BTEX flux and methanogen numbers. For profile
(C), water table elevations fluctuate seasonally over the shaded gray region and methanogens are present only
below the lowest water table elevation.

(BTEX) for the three vertical profiles. The computed flux is the product of the average
water table gradient between Sites A and C (0.0024), the estimated permeabilities, and the
total BTEX concentrations. Values between data points for BTEX and permeability were
estimated by linear interpolation. The correspondence between increased flux and
methanogens numbers is generally quite good. The higher peak flux at Site B is associated
with a higher number of methanogens compared to Site A. Unfortunately, the peak in flux
near the base of Site A (419.7 m) cannot be associated with a peak in methanogens
because no microbial data exist for this location. A Pearson correlation analysis for the 14
methanogen MPN data values from below the oil body together with the computed BTEX
fluxes at the corresponding elevations gives a significant value of 0.53 (n = 14; p < 0.06).
In areas of lower flux, found along the plume fringes and in low permeability zones,
iron-reducer numbers remain high. In these locations, the relatively slow migration of
contaminated water by transverse dispersion and diffusive flux results in a lower
cumulative hydrocarbon flux over time. Thus, the available Fe(III) is less depleted and
these areas still support high populations of iron-reducers compared to nearby locations
with higher hydrocarbon flux. As a consequence, the plume under the oil body still
contains both methanogenic and iron-reducing zones even though this portion of the
aquifer has been anaerobic for many years.
At Site C, the picture is more complex because the location of peak flux is above the
methanogenic zone in the top 50 cm of the saturated zone (Fig. 5c). To understand these
data, we examined the water level changes in well 531 located adjacent to Site C (see Fig.
2 for locations). Between 1985 and the date of the microbial analyses in 1997, the water
levels fluctuated about 0.6 m seasonally. The shaded area in Fig. 5c shows the average
water table over this period plus or minus two standard deviations. Culturable methanogens are found only below the lowest level of the water table. Presumably, the seasonal
presence of oxygen makes locations above the annual low water table toxic to metha-

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

399

nogens. Another indication that changing redox conditions affect the microbial data at this
location is the coexistence of iron reducers and aerobes at both the top and base of the
profile. Dissolved Fe(II) is below detection and the cores were visually more orangecolored at both the top and the base of the profile compared to the center. These
observations suggest that dissolved oxygen is intermittently present at the fringes of the
profile creating conditions that alternately favor each population.
The effects of water table oscillations on the Site C profile also obscure the relationship
of hydrocarbon flux to the microbial zones. During water table low stands, the advective
flux of contaminants through elevations above the water table is negligible whereas the
flux through the positions at or below the water table continues to occur. The 1997 data
correspond to a year in which the water table reached a level of 423.84 m, which is higher
than any value measured in previous years. If flux were adjusted for relative time during
which the water levels lie above the minimum water table, the peak in the effective
hydrocarbon flux plot would lie well below that shown. However, this type of analysis is
not possible with the existing data because seasonal variations in profiles of pore-water
chemistry are not available.
4.3. Modeling of the biogeochemical processes
Essaid et al. (1995) constructed a two-dimensional solute transport model with
biodegradation for the Bemidji, MI, crude oil spill site. The simulations included the
sequential degradation of volatile and nonvolatile fractions of dissolved organic carbon by
aerobic processes, manganese reduction, iron reduction, and methanogenesis. Despite the
considerable uncertainty in the model parameter estimates, results of simulations reproduced the general features of the observed ground-water plume and the measured
microbial concentrations from the time of the spill in 1979 to 1992. To examine the
predicted long-term behavior of the plume and microbial populations, the simulation of
Essaid et al. (1995) was extended from 13 to 20 years.
Fig. 6 shows the simulated concentrations at a location 36 m downgradient from the
center of the oil body for volatile and nonvolatile dissolved organic carbon (VDOC and
NVDOC, respectively); solid phase Mn(IV); solid phase Fe(III); and aerobic, manganese /
iron reducer, and methanogenic biomass. It is interesting to note that the simulations
predict that the VDOC concentration (this includes benzene, toluene, ethylbenzene, m-, pand o-xylene) reaches a minimum at about 11 years following the spill and then increases.
The increase in VDOC concentration corresponds to the depletion of solid phase Fe(III).
The increase in VDOC concentrations as Fe(III) is depleted is in agreement with recent
observations at the site (Cozzarelli et al., 2001).
The manganese /iron (Mn /Fe) reducer biomass concentration shows a rapid increase
during the first 5 years following the spill when manganese reduction is the dominant
process. As Mn(IV) is depleted, the biomass concentration decreases slightly until iron
reduction becomes significant, resulting in increased Mn /Fe reducer growth. The Mn /Fe
reducer biomass reaches a peak at about 12 years and then begins to decrease as
Fe(III) is depleted. The biomass of the methanogenic consortium in the simulations
begins to dominate the population after 16 years. It is important to note that the model
combines the activities of the entire methanogenic consortium into one process. Thus,

400

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

Fig. 6. Simulated concentrations at a location 36 m downgradient from the center of the oil body for (a) volatile
and nonvolatile dissolved organic carbon (VDOC and NVDOC, respectively), (b) solid phase manganese (Mn4+),
solid phase iron (Fe3+), and (c) aerobic, manganese/iron reducer, and methanogen biomass. Simulated
concentrations of VDOC rise slowly between 15 and 20 years as solid phase iron oxides are depleted.

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

401

the overall degradation rate of hydrocarbons to carbon dioxide and methane represents
the rate-limiting step in combined series of fermentation and methanogenesis reactions.
Moreover, the methanogenic biomass in the model reflects the biomass of the total
consortium. The growth of the methanogenic consortium in the simulations does not
appear to be limited by nutrients during the simulated time period. In reality, although
the supply of carbon for growth is plentiful, other required nutrients such as
phosphorus or nitrogen will limit the growth of the methanogens and fermenters in
the aquifer.
The prediction by the simulations that the biomass of the methanogenic consortium will
eventually exceed that of iron reducers correctly reflects the character of the microbial data
in the designated methanogenic zones at sites A and B (Figs. 3 and 4). These two sites are
both located upgradient of the simulated location. The microbial data show that the aquifer
at site C has not yet reached the stage where the methanogenic consortium dominates the
population. Similarly, this site is located downgradient of the simulated location. The
modeling indicates that the methanogenic zone expands downgradient at an average rate of
2.25 m / year. However, this prediction is based on a homogeneous aquifer with average
properties for the Bemidji site. The microbial data indicate that after almost 20 years the
populations of iron reducers and the methanogenic consortium at Site C are almost equal.
Thus, the data from the three vertical profiles together with the widely spaced sample
locations shown in Fig. 3 indicate that a horizontally continuous methanogenic zone
extends downgradient for almost 60 m from the center of the oil body (Bekins et al.,
1999). Thus, the actual rate of horizontal expansion of the methanogenic zone is almost 3
m /year.
The plots in Figs. 4 and 5 suggest that the methanogenic zone occurs along a
connected, high flux pathway in the Bemidji aquifer. This observation is consistent with
previous modeling results of contaminant plumes by Webb and Anderson (1996). They
showed that in subsurface systems with contrasts in hydraulic conductivity, such as
glacial outwash aquifers, the location and magnitude of flow is constrained by the
internal sediment structures. A recent paper by Harvey and Gorelick (2000) argues that
solute transport in a heterogeneous aquifer can be described by advection through
mobile regions together with rate-limited diffusion into small-scale immobile regions.
Our data suggest that this model accurately describes solute transport and biodegradation at the Bemidji site and explains some aspects of the observed microbial distribution.
Within the advecting plume, methanogenic zones evolve first in the mobile regions
where the hydrocarbon flux is highest. The higher flux of contaminants causes a more
rapid decrease in sediment Fe(III) oxides allowing methanogenic conditions to develop
more rapidly in these regions than a strictly homogeneous aquifer model would predict.
In contrast, iron-reducing conditions persist longer than the homogeneous model
predicts in the immobile regions of the aquifer where diffusive mass transfer is rate
limiting.
Aqueous phase electron acceptors may also be depleted first in longitudinally
contiguous high flux pathways. Scholl et al. (1999) demonstrated that dissolved electron
acceptors in an initially uncontaminated aquifer are depleted fastest in high permeability
zones. Because longitudinal dispersivity is usually 100 times greater than transverse
dispersivity (Garabedian et al., 1991), the plume preferentially mixes with uncontami-

402

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

nated water in the principle direction of flow. Thus, the plume grows longitudinally as
electron acceptors are depleted from the water and sediments along the high permeability
horizons. In contrast, lower permeability immobile regions retain supplies of favorable
electron acceptors longer because the biodegradation rate is limited by diffusion of
contaminants into and electron acceptors out of the immobile regions.
4.4. Implications for monitored natural attenuation
Although high flux zones where methanogenic conditions develop are laterally
continuous, they may be very thin vertically, making them difficult to locate. Cozzarelli
et al. (1999) noted that very thin redox zones at a gasoline-contaminated site resulted
in mixed geochemical indicators in well-water samples. At Site C, the methanogenic
niche spans an interval of <1 m but dissolved methane is present over a 4 m interval
(see Fig. 9 in Cozzarelli et al., 2001). This aspect of plume behavior complicates the
use of aqueous geochemical indicators to infer terminal electron accepting processes.
Assays of microbial populations attached to the sediments provide a better indication of
the microbial physiologic processes at a given location than do aqueous redox indicators.
Due to the expense of microbial analyses, it is useful to have a cheaper field parameter
that can be used to locate horizons where microbial populations change. At the Bemidji
site, we found that the methanogenic zones generally correspond to the centers of peak
concentrations in dissolved Fe(II). A strategy of analyzing dissolved iron in the field from
pore waters drained from cores over 15 cm intervals (see Section 3) proved useful in
formulating our 1997 sediment sampling scheme for microbial analyses. Note that the
activities of the microbial populations attached to the sediments as measured by the MPN
method were ultimately used to define the microbial zones. The field Fe(II) concentrations
were simply a successful field proxy for choosing sample locations. At each site, a
different proxy would need to be identified. The key idea is that a search for correlations
between microbial population changes and small-scale variations in either pore-water
chemistry or sediment properties may be useful for formulating a strategy that limits the
number of required microbial analyses.
The development of longitudinally continuous methanogenic zones along high
permeability horizons has implications for risk analysis at sites with large sources of
reduced carbon contaminants. Degradation rates of hydrocarbons under methanogenic
conditions are known to differ from other anaerobic processes (Krumholz et al., 1996).
Thus, as the methanogenic zone expands, breakthrough of compounds that degrade
slowly or resist degradation under methanogenic conditions will occur. The small
population of methanogens found in the center of the plume at Site C indicates that
methanogenic conditions are just beginning there. As expected, there is also breakthrough
of o-xylene at 422.6 423.2 m (see Fig. 11 in Cozzarelli et al., 2001), which is
approximately the same elevation where incipient methanogenic conditions are found
(422.3 422.8 m).
Cozzarelli et al. (2001) discuss how the expansion of the methanogenic zone has
resulted in slow vertical and horizontal growth of the hydrocarbon plume. Nielsen et al.
(1995) observed similar behavior at a landfill site where benzene isomers migrated with

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

403

little degradation through the portion of the aquifer depleted in Fe(III) but degraded
rapidly in the zone where significant Fe(III) is still present. At the Bemidji site,
Geobacter spp. capable of oxidizing benzene under iron-reducing conditions are found
only in the zone where significant Fe(III) is present (Anderson et al., 1998; RooneyVarga et al., 1999) and benzene degrades rapidly in this zone (Anderson and Lovley,
1999).

5. Summary and conclusions

Combined studies of microbial populations together with physical and chemical
conditions in the Bemidji aquifer have provided insights into the long-term behavior of
petroleum hydrocarbon plumes. The dominant physiologic type of the microbial
population changes from iron reducing to methanogenic over vertical scales as small
as 25 cm. Near the oil body, methanogenic zones have evolved in two high-flux areas.
One zone is found within the nonaqueous oil where high fluxes are due to local
dissolution and diffusion of contaminants from the oil. A second zone exists along high
permeability horizons in the laterally migrating contaminant plume. At a distance of 65
m downgradient from the oil, lower numbers of methanogens indicate that methanogenic
conditions are less advanced than below the oil. The spatial correspondence between
methanogenic zones and peaks in hydrocarbon flux suggest that the methanogenic zone
in the plume is growing laterally along a narrow, connected, high permeability pathway.
Results presented by Cozzarelli et al. (2001) show that the growth of the methanogenic
zone results in a slow expansion of the hydrocarbon plume. A computer simulation of
the plume predicted an initial decrease in concentrations between 5 and 10 years after
the spill at an observation well 36 m downgradient of the oil body, followed by an
increase in simulated concentrations between 10 and 15 years as iron oxides are depleted
and iron-reducing bacteria decrease in number. At sites with large petroleum hydrocarbon plumes from persistent nonaqueous sources, risk analyses should account for
expansion of methanogenic conditions along narrow, connected high permeability
pathways.

References
Adrian, N.R., Robinson, J.A., Sulflita, J.M., 1994. Spatial variability in biodegradation rates as evidenced by
methane production from and aquifer. Applied and Environmental Microbiology 60 (10), 3632 3639.
Anderson, R.T., Lovley, D.R., 1999. Naphthalene and benzene degradation under Fe(III)-reducing conditions in
petroleum contaminated aquifers. Bioremediation Journal 3 (2), 121 135.
Anderson, R.T., Rooney-Varga, J.N., Gaw, C.V., Lovley, D.R., 1998. Anaerobic benzene oxidation in the
Fe(III) reduction zone of petroleum-contaminated aquifers. Environmental Science and Technology 32 (9),
1222 1229.
Baedecker, M.J., Cozzarelli, I.M., Eganhouse, R.P., Siegel, D.I., Bennett, P.C., 1993. Crude oil in a shallow sand
and gravel aquifer: III. Biogeochemical reactions and mass balance modeling in anoxic groundwater. Applied
Geochemistry 8, 569 586.
Baedecker, M.J., Cozzarelli, I.M., Bennett, P.C., Eganhouse, R.P., Hult, M.F., 1996. Evolution of the contaminant
plume in an aquifer contaminated with crude oil, Bemidji, Minnesota. In: Morganwalp, D.W., Aronson, D.A.

404

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

(Eds.), US Geological Survey Toxic Substances Hydrology Program Proceedings of the Technical Meeting,
Colorado Springs, Colorado, September 20 24, 1993, pp. 613 620, US Geological Survey WaterResources Investigations Report 94-4015.
Bekins, B.A., Godsy, E.M., Warren, E., 1999. Distribution of microbial physiologic types in an aquifer contaminated by crude oil. Microbial Ecology 37, 263 275.
Bennett, P.C., Siegel, D.E., Baedecker, M.J., Hult, M.F., 1993. Crude oil in a shallow sand and gravel aquifer: I.
Hydrogeology and inorganic geochemistry. Applied Geochemistry 8, 529 549.
Cozzarelli, I.M., Baedecker, M.J., Eganhouse, R.P., 1990. Transformation of monoaromatic hydrocarbons to
organic acids in anoxic groundwater environment. Environmental Geology and Water Science 16 (2),
135 141.
Cozzarelli, I.M., Baedecker, M.J., Eganhouse, R.P., Goerlitz, D.F., 1994. The geochemical evolution of lowmolecular-weight organic acids derived from the degradation of petroleum contaminants in groundwater.
Geochimica et Cosmochimica Acta 58, 863 877.
Cozzarelli, I.M., Baedecker, M.J., Aiken, G.R., Phinney, C., 1996. Small-scale chemical heterogeneities in a
crude-oil contaminated aquifer, Bemidji, Minnesota. In: Morganwalp, D.W., Aronson, D.A. (Eds.), US Geological Survey Toxic Substances Hydrology Program Proceedings of the Technical Meeting, Colorado
Springs, Colorado, September 20 24, 1993, pp. 647 652, US Geological Survey Water-Resources
Investigations Report 94-4015.
Cozzarelli, I.M., Herman, J.S., Baedecker, M.J., Fischer, J.M., 1999. Geochemical heterogeneity of a gasolinecontaminated aquifer. Journal of Contaminant Hydrology 40 (3), 261 284.
Cozzarelli, I.M., et al., 2001. Progression of natural attenuation processes at a crude-oil spill site: I. Geochemical
evolution of the plume. Journal of Contaminant Hydrology 53, 369 386.
Dillard, L.A., Essaid, H.I., Herkelrath, W.N., 1997. Multiphase flow modeling of a crude-oil spill site with a
bimodal permeability distribution. Water Resources Research 33, 1617 1632.
Eganhouse, R.P., et al., 1993. Crude oil in a shallow sand and gravel aquifer: II. Organic geochemistry. Applied
Geochemistry 8, 551 567.
Eganhouse, R.P., Dorsey, T.F., Phinney, C.S., Westcott, A.M., 1996. Processes affecting the fate of monoaromatic
hydrocarbons in an aquifer contaminated by crude oil. Environmental Science and Technology 30 (11),
3304 3312.
Essaid, H.I., et al., 1995. Simulation of aerobic and anaerobic biodegradation processes at a crude oil spill site.
Water Resources Research 31 (12), 3309 3327.
Franzi, D.A., 1988. Surficial and subsurface distribution of aquifer sediments at the Bemidji, Minnesota, research
site. In: Ragone, S.E. (Ed.), US Geological Survey Program on Toxic Waste-Groundwater Contamination,
Proceedings of the Second Technical Meeting, Cape Cod, Massachusetts, October 21 25, 1985, pp. c5 c10,
US Geological Survey Open File Report 86-481.
Garabedian, S.P., LeBlanc, D.R., Gelhar, L.W., Celia, M.A., 1991. Large-scale natural gradient tracer test in sand
and gravel, Cape Cod, Massachusetts: 2. Analysis of spatial moments for a nonreactive tracer. Water Resources Research 27 (5), 911 924.
Godsy, E.M., 1980. Isolation of Methanobacterium bryantii from a deep aquifer by using a novel broth-antibiotic
disk method. Applied and Environmental Microbiology 39 (5), 1074 1075.
Godsy, E.M., Goerlitz, D.F., Grbic-Galic, D., 1992. Methanogenic biodegradation of creosote contaminants in
natural and simulated ground water ecosystems. Ground Water 30 (2), 232 242.
Harvey, C., Gorelick, S.M., 2000. Rate-limited mass transfer or macrodispersion: which dominates plume evolution at the Macrodispersion Experiment (MADE) site? Water Resources Research 36 (3), 637 650.
Herkelrath, W.N., 1999. Impacts of remediation at the Bemidji oil spill site. In: Morganwalp, D.W., Buxton, H.T.
(Eds.), US Geological Survey Toxic Substances Hydrology ProgramProceedings of the Technical Meeting,
Charleston, South Carolina, March 8 12, 1999, Volume 3, Subsurface Contamination from Point Sources,
pp. 195 200, US Geological Survey Water-Resources Investigations Report 99-4018C.
Hess, K.M., Herkelrath, W.N., Essaid, H.I., 1992. Determination of subsurface fluid contents at a crude-oil spill
site. Journal of Contaminant Hydrology 10, 75 96.
Hult, M.F., 1984. Groundwater contamination by crude oil at the Bemidji, Minnesota, research site an introduction. In: Hult, M.F. (Ed.), Groundwater Contamination by Crude Oil at the Bemidji, Minnesota, Research
Site, pp. 1 15, US Geological Survey Water-Resources Investigations Report 84-4188.

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

405

Krumbein, W.C., Monk, G.D., 1942. Permeability as a function of the size parameters of unconsolidated sand.
Transactions of the American Institute of Mining and Metallurgical Engineering 151, 153 163.
Krumholz, L.R., Caldwell, M.E., Sulflita, J.M., 1996. Biodegradation of BTEX hydrocarbons under anaerobic
conditions. In: Crawford, R.L., Crawford, D.L. (Eds.), Bioremediation: Principles and Applications. Cambridge Univ. Press, pp. 61 99.
Lovley, D.R., Phillips, E.J.P., 1986. Organic matter mineralization with reduction of ferric iron in anaerobic
sediments. Applied and Environmental Microbiology 51 (4), 683 689.
Lovley, D.R., et al., 1989. Oxidation of aromatic compounds coupled to microbial iron reduction. Nature 339,
297 299.
Lovley, D.R., Woodward, J.C., Chapelle, F.H., 1994. Stimulated anoxic biodegradation of aromatic hydrocarbons
using Fe(III) ligands. Nature 370, 128 131.
Morgan, J.W., Evison, L.M., Forster, C.F., 1991. Changes to the microbial ecology in anaerobic digesters treating
ice cream wastewater during start-up. Water Resources 25 (6), 639 653.
Murphy, F., Herkelrath, W.N., 1996. A sample-freezing drive shoe for a wire line piston core sampler. Ground
Water Monitoring and Remediation 16, 86 90.
Nielsen, P.H., Christensen, T.H., 1994. Variability of biological degradation of aromatic hydrocarbons in an
aerobic aquifer determined by laboratory batch experiments. Journal of Contaminant Hydrology 15, 305
320.
Nielsen, P.H., Bjarnadottir, H., Winter, P.L., Christensen, T.H., 1995. In situ and laboratory studies on the fate of
specific organic compounds in an anaerobic leachate plume: II. Fate of aromatic and chlorinated aliphatic
compounds. Journal of Contaminant Hydrology 20, 51 66.
NRC, 1993. In Situ Bioremediation National Academy Press, Washington, DC, 207 pp.
Postma, D., 1993. The reactivity of iron oxides in sediments: a kinetic approach. Geochimica et Cosmochimica
Acta 57, 5027 5034.
Roden, E.E., Zachara, J.M., 1996. Microbial reduction of crystalline iron(III) oxides: influence of oxide surface
area and potential for cell growth. Environmental Science and Technology 30 (5), 1618 1628.
Rooney-Varga, J.N., Anderson, R.T., Fraga, J.L., Ringelberg, D., Lovley, D.R., 1999. Microbial communities
associated with anaerobic benzene degradation in a petroleum-contaminated aquifer. Applied and Environmental Microbiology 65 (7), 3056 3063.
Schink, B., 1997. Energetics of syntrophic cooperation in methanogenic degradation. Microbiology and Molecular Biology Reviews 61 (2), 262 280.
Scholl, M.A., Cozzarelli, I.M., Christenson, S.C., Breit, G.N., Schlottmann, J.L., 1999. Aquifer heterogeneity at
the Norman, Oklahoma, landfill and its effect on observations of biodegradation processes. In: Morganwalp, D.W., Buxton, H.T. (Eds.), US Geological Survey Toxic Substances Hydrology ProgramProceedings of the Technical Meeting, Charleston, South Carolina, March 8 12, 1999, Volume 3, Subsurface
Contamination from Point Sources, pp. 557 567, US Geological Survey Water-Resources Investigations
Report 99-4018C.
Smith, R.L., Harvey, R.W., LeBlanc, D.R., 1991. Importance of closely spaced vertical sampling in delineating
chemical and microbiological gradients in groundwater studies. Journal of Contaminant Hydrology 7, 285
300.
Tuccillo, M.E., Cozzarelli, I.M., Herman, J.S., 1999. Iron reduction in the sediments of a hydrocarbon-contaminated aquifer. Applied Geochemistry 4 (5), 71 83.
Urritia, M.M., Roden, E.E., Fredrickson, J.K., Zachara, J.M., 1998. Microbial and surface chemistry controls on
reduction of synthetic Fe(III) oxide minerals by the dissimilatory iron-reducing bacterium Shewanella alga.
Geomicrobiology 15, 269 291.
Urritia, M.M., Roden, E.E., Zachara, J.M., 1999. Influence of aqueous and solid-phase Fe(II) complexants on
microbial reduction of crystalline iron(III) oxides. Environmental Science and Technology 33 (22), 4022
4028.
Warren, E., Bekins, B.A., Godsy, E.M., 1999. Inhibition of acetoclastic methanogenesis by crude oil from
Bemidji, Minnesota. In: Morganwalp, D.W., Buxton, H.T. (Eds.), US Geological Survey Toxic Substances
Hydrology ProgramProceedings of the Technical Meeting, Charleston, South Carolina, March 8 12, 1999,
Volume 3, Subsurface Contamination from Point Sources, pp. 223 230, US Geological Survey WaterResources Investigations Report 99-4018C.

406

B.A. Bekins et al. / Journal of Contaminant Hydrology 53 (2001) 387406

Webb, E.K., Anderson, M.P., 1996. Simulation of preferential flow in three-dimensional, heterogeneous conductivity fields with realistic internal architecture. Water Resources Research 32 (3), 533 545.
Yoon, W.B., Rosson, R.A., 1990. Improved method of enumeration of attached bacteria for study of fluctuation in
the abundance of attached and free-living bacteria in response to diel variation in seawater turbidity. Applied
and Environmental Microbiology 56 (3), 595 600.