Analysis of matrix effects critical to microbial transport in organic waste-affected soils across laboratory and field scales



[1] Organic waste applications to soil (manure, various wastewaters, and biosolids) are among the most significant sources of bacterial contamination in surface and groundwater. Transport of bacteria through the vadose zone depends on flow path geometry and stability and is mitigated by interaction between soil, soil solution, air-water interfaces, and characteristics of microbial surfaces. After initial entry, the transport through soil depends on continued entrainment of bacteria and resuspension of those retained in the porous structure. We evaluated the retention of bacteria-sized artificial microspheres, varying in diameter and surface charge and applied in different suspending solutions, by a range of sieved soils contained in minicolumns, the transport of hydrophobic bacteria-sized microspheres through undisturbed soil columns as affected by waste type under simulated rainfall, and the field-scale transport of Enterococcus spp. to an unconfined sandy aquifer after the application of liquid manure. Microsphere retention reflected microsphere properties. The soil type and suspending solution affected retention of hydrophilic but not hydrophobic particles. Retention was not necessarily facilitated by manure-microsphere-soil interactions but by manure-soil interactions. Undisturbed column studies confirmed the governing role of waste type on vadose-zone microsphere transport. Filtration theory applied as an integrated analysis of transport across length scales showed that effective collision efficiency depended on the distance of travel. It followed a power law behavior with the power coefficient varying from ∼0.4 over short distances to >0.9 over 1 m (i.e., very little filtration for a finite fraction of biocolloids), consistent with reduced influence of soil solution and biocolloid properties at longer travel distances.

1. Introduction

[2] The transport of bacteria through soils has long been recognized as an important phenomenon for the contamination of water resources [e.g., Culley and Phillips, 1982; Goss et al., 1998; Joy et al., 1998; Patni et al., 1985; Shreshta et al., 1997; Unc and Goss, 2003]. Many researchers have investigated the properties of bacteria, which are important for their retention to surfaces, and hence the potential for their transport in the liquid phase [e.g., Lachica, 1990; van Loosdrecht et al., 1987]. A common transport pathway for fecal bacteria into soils is through the application of organic wastes, such as animal manure, sewage biosolids, or food (processing) wastes and the subsequent infiltration into and percolation through soils.

[3] Bradford and Torkzaban [2008] concluded from a review of the literature that the essential parameters governing the transport of colloids through unsaturated soils are the thermodynamic interactions mediated by surface charge properties and solution chemistry, soil water content, pore geometry and roughness of soil internal surfaces, colloid concentration, and individual particle size. Experimental evidence shows that air-water interfaces can also be relevant in the transport of particles under unsaturated conditions [DeNovio et al., 2004; Schäfer et al., 1998a, 1998b]. However, they may be less significant where charged surfaces are present, such as clay minerals and organic matter [Chu et al., 2001, 2003]. The deposition of colloids onto soil surfaces is enhanced when they are present at large concentration, and the newly deposited material will act in turn as a retention surface accelerating removal from the transporting fluid and the clogging of pores [Zhang et al., 2010].

[4] Electrochemical interactions play an essential role in the initial attachment of charged particles, including bacteria, to surfaces. Elevated ionic concentrations in the soil solution due to charged organic groups and free ions can favor electrostatic interactions by reducing the forces of repulsion [Jones, 1975] and by increasing the surface tension of water [Reid et al., 1991]. Free ions also compete for retention sites with mobile charged surfaces [Makin and Beveridge, 1996]. These may affect the extent of the hydrophobic or hydrophilic behavior of charged particles [Vargha-Butler et al., 1985] and their interaction with any air-water interface [Powelson and Mills, 1996].

[5] The surface properties of bacterial cells vary according to the particular expression of diverse gene sets under changeable environmental conditions. Therefore, assessment of the importance of particular cell properties for bacterial transport under complex environmental conditions is difficult. Artificial microspheres of known properties have been useful for assessing the transport of bacteria-sized hydrophilic or hydrophobic particles through saturated aquifer sediments [Cumbie and McKay, 1999; Harvey et al., 1993, 1995; Lindqvist and Bengtsson, 1995]. Passmore et al. [2010] showed that microspheres with comparable density, size, and surface properties to bacteria are also useful as experimental surrogates for pathogens in evaluating transport through unsaturated soil. In most of these studies the organic carbon properties of the suspending solutions varied little.

[6] There has been limited research into the significance of organic carbon components of the soil solution or of the properties of cell surfaces in microbial transport through the vadose zone after an application of organic wastes. Increases in pH were correlated with enhanced attachment of Bacillus spores to silt [Kim et al., 2009] but decreased the adhesion of Pseudomonas putida to clay minerals or goethite [Jiang et al., 2007]. Theoretically, electrostatic adhesion of bacterial cells to charged soil minerals depends on the pH of the suspending solution and the isoelectric points of the two interacting surfaces [Rong et al., 2010]. Other charged colloids originating in soil or organic waste and having variable isoelectric points can mediate these interactions. This pool of particles with variable charges under the same pH makes a prediction of the role of pH in agricultural soils under naturally occurring conditions more complex. It is notable that soil pores available to transport microbes, more common in well-structured soils [e.g., Artz et al., 2005], are coated with polymeric organic carbon compounds [Kaiser and Guggenberger, 2003] that obstruct soil mineral surfaces. Thus, although relatively simple in concept the role of the pH in microbial transport through natural soils is still unclear.

[7] According to the double-layer theory of electrostatic interaction [Derjaguin and Landau, 1941; Verwey and Overbeek, 1948], a large ionic strength and a predominance of polyvalent ions would be expected to enhance microbial adhesion. Small ionic strength has been shown to enhance bacterial transport [Li and Logan, 1999]. In the context of a net negative charge, which is common in most soils, the presence of cations in the solution, especially polyvalent cations, is expected to lower the attachment energy and thus limit transport. Nevertheless, the proportion of polyvalent cations may not be relevant at ionic strengths >0.1 M [Wasserman and Felmy, 1998]. Moreover, at the typical concentrations of cells and similar colloids the deposition rates on the surface may overwhelm the available surface [e.g., Guber et al., 2009], and thus allow essentially equivalent transport rates independent of the ionic strength [Schinner et al., 2010] making retention a function of transport path length.

[8] The detachment from soil surfaces of motile, flagellated bacteria is enhanced at more biologically optimal temperatures [McCaulou et al., 1995]. However, self-motility of bacteria is not of significance for long distance transport [Gannon et al., 1991]. Temperature mostly affects the pool of viable microbial cells available for transport. The general agreement is that cooler temperatures maintain the size of this pool and that warmer conditions deplete this pool through die-off, competition, and predation [Jamieson et al., 2002]. Nevertheless, this is true to varying degrees with diverse organisms and diverse environmental and land management conditions [Brennan et al., 2010; Unc et al., 2004]. While thermodynamic considerations indicate chemical and biological activities vary according to temperatures, there is no information suggesting this to be of relevance for the temperature ranges common in soils when freezing is not occurring, the conditions prevailing during our tests.

[9] Suspended organic matter colloids may enhance microbial transport through soil by offering attachment surfaces independent of the fixed soil surfaces. Experimental evidence shows that suspended manure components enhance the transport of associated bacteria [Unc et al., 2004] by minimizing their attachment to soil [Guber et al., 2007, 2009].

[10] It has been generally accepted that microbial transport through the vadose zone is mostly passive and thus governed by water fluxes. Waste type has been shown to affect bulk vadose-zone water flow [Mosaddeghi et al., 2009; Unc and Goss, 2003]. More specifically, the retention of water in the soil matrix and flow kinetics in the macropores are modified and consequently the water partitioning into infiltration or surface runoff is modified [Unc and Goss, 2006]. Partitioning of the vadose water fluxes between the matrix and macropore flow and associated microbial filtration coefficients was shown to depend on the soil water saturation status [Harter et al., 2008; Mosaddeghi et al., 2010].

[11] Integrated macroscale studies may suggest that variability in soil parameters and, critically, soil structural parameters [Pang et al., 2008] and associated soil water regimes [Lin et al., 1999; Torkzaban et al., 2008], as determined by soil chemical and mineralogical properties or induced by management practices [Bronick and Lal, 2005], may be the most relevant parameters governing the vadose zone transfer of microbial contaminants. This begs the question if, in a system as chemically, physically, and biologically complex as is a soil amended with organic wastes, porescale level understanding of colloid transport can be effectively scaled to risk assessment exercises at field and landscape scale.

[12] The objective of this integrated study was to evaluate the extent to which the understanding of proximate causes obtained at various scales may be combined to assess the potential for bacterial transport through the vadose zone following land application of organic waste. This incorporated soil as a transfer matrix and organic waste as a modifier. The approach amounts to the multiple-scale evaluation of the significance of soil type as affected by the interactions with various types of bacteria-bearing organic waste for microbial retention and transport through the vadose zone. The first experiment involved the use of fully saturated soil minicolumns to determine the attenuation of batches of fluorescent microspheres, of various charge characteristics. Microspheres, suspended in manure liquors containing organic material of different sizes or distilled water, were applied to the soil columns, which were leached using aliquots of the same liquids as the suspending solution. A second experiment evaluated the significance of soil and waste types for the transport of hydrophobic microspheres through larger-volume undisturbed soil columns under unsaturated conditions simulating real intensity recurring rains. The final experiment, which investigated microbial transport to a shallow unconfined alluvial aquifer resulting from manure applied by irrigation, allowed assessing whether observations made with microbial colloids at the field scale were consistent with those made at the smaller scales. The various experimental approaches required the use of standard statistics, common to manipulative experiments, and exploratory statistics common to natural or observational experiments. Integrative empirical equations were eventually employed to assess the stability of the results obtained from these discrete experiments across multiple scales (Figure 1).

Figure 1.

Schematics of the critical parameters for the three vadose-zone transport experimental scales.

2. Materials and Methods

2.1. Soil Minicolumn Retention of Microspheres

2.1.1. Soils and Suspending Solutions

[13] Soil from the top 10 cm of the arable layer from three gray-brown podzols (Hapludalfs) from southern Ontario was used (Table 1). A sample from each of the soils was air-dried, passed through a 2 mm sieve, and then again through a 1 mm sieve. The 1–2 mm fraction was retained for the investigation.

Table 1. Soils Used in the Minicolumns Retention Testsa
Soil TypeTexture (% Weight)Organic Matter (% Weight)CEC (cmolc+ kg−1)pH in Water
  • a

    Soil properties have been measured by the Laboratory Services at the University of Guelph. Standard methodology was employed [Carter, 1993].

Fox sandy loam632982.212.988.0
Conestogo silt loam3050204.827.197.5
Haldimand silty clay1148415.628.547.2

[14] Distilled water (“ddwater”) and two manure-filtrate “solutions” were used as suspending fluids. Liquid swine manure, obtained from prestorage tanks at the Arkell Research Station of the University of Guelph, with a dry matter content of ∼1%, was centrifuged for 20 min at 23,000 × g at 20°C. The decanted supernatant was passed through a 0.45 μm cellulose acetate filter. The resulting filtrate, the “<0.45 micron solution,” had a dry matter content of ∼0.03%; it was stored in the dark in amber-colored bottles at 4°C and used within 24 h. The second manure filtrate was obtained from fresh solid cattle manure, collected from the Elora Research Station of the University of Guelph. Manure liquor, which drained freely from the bottom of a barn-stored solid manure pile was collected. This liquid was centrifuged at 3000 × g for 10 min, and passed through a 1 μm Whatmann filter. This resulting filtrate, the “<1 micron solution,” had a final dry matter content of ∼0.3% and suspended matter was visibly present. Batches of this manure filtrate were stored at 4°C and used within 4 d. Fresh filtrate samples were prepared routinely from the same initial collection of manure that was stored in the dark at 4°C. Both filtrates had a pH of 8.6 and an electrical conductivity (EC) of ∼9 mS cm−1.

2.1.2. Artificial Microspheres

[15] Fluorescent microspheres of known surface charge and diameter (Molecular Probes, Inc. and Polysciences, Inc.) were selected to have a size similar to bacterial cells but having a wide range of surface charges (Table 2). They had a density of 1.05 g cm−3, similar to that of bacterial cells. The noncharged (hydrophobic) microspheres were made of latex, while the charged ones were made of polystyrene with negatively charged carboxylic groups cross-linked on their surface. For well-defined surfaces the effective hydrophobicity of a particle is inversely related to its surface charge [Busscher et al., 1984].

Table 2. Characteristics of the Artificial Microspheres Used in the Minicolumn Retention Tests
TypeDiameter (μm)Surface Charge (meq g−1 DM)Reaction in WaterCharge Density (meq μm−2)Peak Excitation Wavelength (nm)Peak Emission Wavelength (nm)

[16] To enumerate the fluorescent microspheres in leachate from soil columns, samples were first diluted and an aliquot filtered through a 0.45 μm filter with an active diameter of 1.5 cm. The filters were mounted on a microscope slide covered with a glass cover slip, which was fixed by glue at the corners. Two replicate 1 mL dilutions (100 and 10−1) were prepared from each filtrate sample. This procedure resulted in a distribution of microspheres on the slides that was satisfactory for counting. Filters were stored in the dark and examined within 1 week using a confocal fluorescent microscope equipped with a computerized image collection system. From each filter set prepared from the two dilutions one was selected for image collection (i.e., if the 100 dilution was unreadable due to excess microspheres the 10−1 dilution was used). Counts of the number of microspheres present in one microscope field of view varied between samples but varied little within each sample. Thus, three images were randomly taken from each selected filter and saved in electronic format as tagged image format files (tiff). Counts of the number of fluorescent particles, their individual area, and light emission intensity were assessed (Scion Image). The apparent area and emission intensity for the same type of microspheres varied between samples, and frequent recalibration of the microscope's image contrast parameter was necessary to eliminate false counts from fluorescing organic or inorganic soil materials that could interfere with quantification. Thus, the apparent area of an individual fluorescent particle was estimated for each sample. The median value of the area of individual particle in each image was considered to represent the apparent area of an individual microsphere and was used to establish the number of individual particles. Occasionally, the images of some microspheres overlapped. Therefore, any particle that appeared to be larger than the apparent size of one particle must have come from overlapping images. The areas measured for these images were divided by the apparent area of an individual microsphere and the resulting numbers were rounded up to give the number of agglomerated particles. Finally, these values were added to the number of individual microspheres to obtain the total number of microspheres per image. The number of agglomerated particles was expressed as a percent of the total number of particles present, and this value was used in the analyses of the agglomeration of microspheres in the leachate. Agglomeration of microspheres was also used to assess the ability of the leaching solution characteristics to influence it and as a covariate factor.

2.1.3. Minicolumn Experimental Procedure

[17] One gram of the sieved air-dry soils was added to borosilicate glass tubes (0.5 cm diameter and 7 cm length) fitted at the bottom with a fritted plate (pore sizes from 20 to 40 μm). Soils were saturated by slowly immersing the columns in the appropriate solution (ddwater, <0.45 or <1 micron solution). Next, the soils were washed five times with 2 mL of the same suspending solution to achieve electrical equilibrium. Microspheres were washed in ddwater and sonicated prior to use. A microsphere suspension aliquot of 0.1 mL (2.4 × 107 microspheres) was added to each soil column. The suspensions were allowed to equilibrate with the soil surfaces for 15 min. Columns were then leached with a total of 10 mL (∼27.5 pore volumes, pv) added in five successive 2 mL volumes (∼5.5 pv) of the appropriate solution. All measurements were made at 21°C (room temperature). Preliminary testing indicated that no significant movement of microspheres from the columns took place after collection of those in the first 4–8 mL (∼11 pv to ∼22 pv) of any suspending solution. Leaching through the silty clay soil was very slow and required a sample collection to continue for up to 5 h. The difference between the number of microspheres in the material added to a column (based on the concentration) and the total collected in leachate was considered to represent the number retained by the soil, and thus represented the microsphere retention capacity of the column.

2.1.4. Statistical Analysis

[18] The experimental design was completely randomized with soil type, suspending solution type, and microsphere type as factors. The impact of factors on microsphere retention and agglomeration were assessed with a general linear model. The significance of the means was assessed with a Tukey test with a probability (p) of 0.05. Orthogonal contrasts were used to compare the retention of the different types of microspheres for each type of suspending solution. Next, the predictive potential for the microspheres retention and agglomeration, the two response variables, was evaluated for all known characteristics of the soil (percentages of silt, sand, clay, and pH, cation exchange capacity (CEC), organic matter (OM)), microspheres (charge, diameter, and charge density) and manure (dry matter content). As there were large numbers of possible predictors, many of which were correlated with one or more of the other components (multicollinearity), a partial least square (PLS) regression procedure was employed. This is an exploratory procedure similar to principal components analysis. However, while the principal components analysis reflects the covariance structure between the predictor variables, the partial least squares regression reflects the covariance structure between the predictor and response variables. The PLS procedure may be used for theory confirmation and to suggest where relationships might or might not exist [Wold, 1985]. It uses an iterative estimation technique that consists of a series of ordinary least squares analyses, and thus it does not presume any distributional form for measured variables [Hoskuldsson, 1988; Wold, 1985]. Each PLS component is a unique least square regression model that uses linear combinations of the predictor variables (factors). The significance of each predictor (factor) within the first two PLS components are presented in PLS loading graphs. PLS tests were carried out on observations with all of the microspheres and also separately on data subsets containing the observations obtained with the hydrophobic microspheres (A and B) or with the hydrophilic microspheres (C, D, E, and F) (see Table 2). Genstat 11 was used for the analysis of variance while the PLS tests were carried out using the statistical package Minitab 14.

2.2. Leaching of Microspheres Through Undisturbed Soil Columns

2.2.1. Soils and Waste

[19] Cylindrical soil columns, 50 cm in depth and 45 cm in diameter, were collected as undisturbed lysimeters from three different gray-brown podzols (Hapludalfs) at sites across southern Ontario (Table 3). Each lysimeter was amended with treated municipal sewage sludge, either in liquid (2.8% dry matter, “LB”) or dewatered form (29% dry matter, “DW”) at rates equivalent to 5 t dry matter ha−1. Three repeats for each soil-sewage sludge combination were used with the exception of the clay amended with liquid sludge where only two soil cores were used. Fluorescent hydrophobic microspheres of 3.7 μm in diameter (peak excitation wavelength of 505 nm and peak emission wavelength of 515 nm, Polysciences, Inc.) were added to the sludge immediately prior to its application (1.8 × 108 microspheres/soil core). Sludge was worked into the top 2 cm of the soil cores using a spatula.

Table 3. Properties of the Soils Used for the Undisturbed Columns Leaching Testsa
SoilDepth (cm)OM (%)CEC (cmolc+ kg−1)pH in WaterAvailable Ca (mg kg−1 Dry Soil)Soil Texture
  • a

    Soil properties have been measured by the Laboratory Services at the University of Guelph. Standard methodology was employed [Carter, 1993].

Laplaine clay, (calcareous subsoil) (C)0–
Perth clay loam (CL)0– loam
10–203.326.77.55497Silt loam
30–400.624.57.85113Silty clay loam
40–500.522.57.95690Silty clay loam
Fox sandy loam (SL)0–102.513.97.22531Sandy loam
10– loam
20–300.910.67.31510Sandy loam
30–400.78.87.61425Loamy coarse sand
40–500.39.08.04177Fine sandy loam

2.2.2. Irrigation

[20] Soil cores were irrigated with ddwater using a rainfall simulator equipped with 780 endodermic needles connected to a peristaltic roller pump. An initial irrigation event occurred before the application of sludge to insure uniform water content at the start of the experiment. Irrigation was added in further over four discrete events spaced 2–3 weeks apart, with the first event taking place 30 min after the application of sludge. The average water content at the start of a leaching investigation in the different columns and for different irrigation events varied as a function of the natural variability in soil texture and structure. Volumetric water content was monitored automatically using time domain reflectometry probes [Ferré and Topp, 2002] starting at 5 cm and at every subsequent 10-cm depth interval. Initial values of water content at a depth of 5 cm in the soil columns were 33.4% ± 1.9% for clay, 21.7% ± 2.3% for clay loam, and 26.5% ± 0.8% for the sandy loam columns. At the 45 cm depth the initial water content was 38.7% ± 4.2% for clay, 24.4% ± 3.9% for clay loam, and 22.0% ± 2.5% for sandy loam columns. No positive or negative pressure was applied at either boundary and the irrigation did not result in any significant ponding. At each irrigation event ddwater was added at a rate of 10 mm h−1 for 2 h (3000 mL/core), except for the last event when the intensity was doubled (20 mm h−1, 6000 mL per core). Irrigation volumes and intensities were chosen to reflect realistic large rainfall events for southern Ontario or irrigation events in semiarid agricultural regions. Water draining from each column was collected for up to 0.5 h after irrigation had ceased, so that each sampling event lasted a total of 2.5 h.

2.2.3. Sampling and Analysis

[21] Drainage was sampled via a grid collector installed at the core's seepage plane. This allowed the collection of drainage through 64 discrete seepage plane areas for each core. Note that not all of these sampling areas participated in drainage [see Goss et al., 2010]. One sample was collected for each contributing seepage plane area at each irrigation event. Concentration of the microspheres in the drainage water was estimated as described above.

2.3. Field-Scale Transport of Fecal Indicator Enterococcus spp. to Groundwater Under Irrigated Manure Application

2.3.1. Fecal Indicator

[22] Enterococci are Gram-positive bacteria present in the gastrointestinal tract of humans and other mammals. Cells are shed in large numbers (102–106 colony forming units [cfu] g−1) in cattle waste [Unc and Goss, 2004] and are indicator organisms for fecal contamination of water. Cells of Enterococcus spp. are weakly hydrophilic with a water contact angle between 29° and 40° [Gallardo-Moreno et al., 2003; Van Merode et al., 2006a, 2006b] and a ζ-potential (surface charge) between −25 and −40 mV [Cail and Hochella, 2005; Gallardo-Moreno et al., 2003; Schinner et al., 2010; Van Merode et al., 2006a, 2006b]. They are spherical in shape with a median diameter of 0.84 μm and ranging between 0.5 and 1.2 μm [Schinner et al., 2010], and thus are similar in shape, albeit somewhat smaller in size, than the microspheres used in this study. Their density is estimated to be 1.11 g cm−3 [Higgins et al., 1990].

2.3.2. Border Flood Irrigation With Manure Slurry

[23] A border flood irrigation with manure slurry of a forage crop field (Table 4) was monitored at a dairy in the San Joaquin Valley, California for the occurrence of wild Enterococcus spp. in groundwater (Figure 1). Soils at the site included Oakdale sandy loam (Mollic Haploxeralf) and Dinuba sandy loam (Typic Haploxeralf) with coarse loamy sand texture and <1% organic matter in the surface horizon. Soil samples taken from the field site had a near neutral pH (pH 6.9) and contained 12% clay, 18% silt, 70% sand, and 1.44% organic matter. Soils in the area typically contained 65%–80% sand and had a cation exchange capacity of 7.5–10 cmolc+ kg−1 (D. E. Beaudette and A. T. O'Geen, California online soil survey browser, available at, 2011). Well logs showed that the surficial unsaturated alluvial sediments (3–5 m thick) consisted of well-graded, very fine to coarse sand with poorly to moderately graded interbedded silty sand. Groundwater table levels were at 4.5 m below surface with a 0.1% gradient across the field. Estimated hydraulic conductivity in the shallow-most portion of the unconfined aquifer was 18–155 m d−1 based on slug tests in monitoring wells 5.2 cm in diameter with 4.5 m of screen immediately below the water table [Davis, 1995]. Five studies compiled by John and Rose [2005] found that the inactivation rates of Enterococcus, spp. at the groundwater temperature measured during monitoring (∼20°C), range between 0.01–0.8 log10 d−1 with a geometric mean of 0.1 log10 d−1.

Table 4. Concentrations of Enterococcus spp., Electrical Conductivity, and Chloride Ions in Manure Slurry and Background Samples of Groundwater at Site Used for Field-Scale Investigation
MaterialpHEnterococcus spp. (cfu/100 mL)Electrical Conductivity (mS m−1)Chloride (mg L−1)
Manure slurry7.43 × 1066795258.95

[24] Irrigation of the 8 ha fallow field was conducted with undiluted manure slurry between thr harvest of the summer crop (corn, Zea mays) and prior to planting of the winter crop (grains). The field was subdivided into eight blocks for flood irrigation, each 0.9–1.6 ha in extent. Irrigation of each block took between 4 and 7 h with flooding occurring to a depth of ∼0.20 m. Irrigation was interrupted for 2 d midway through the application.

[25] Composite samples (1 L) of the manure slurry were collected directly from outlet valves during irrigation. Large samples (10 L) of groundwater were collected from a monitoring well immediately down gradient of the irrigated field after purging with at least three well volumes. Composite soil samples were collected from eight locations scattered across the field prior to and 1 week following irrigation. Soil samples were homogenized over depth intervals of 0–0.3 m and 0.3–0.6 m below ground level. Additional irrigation water and groundwater samples were also collected for water chemistry analysis. The manure slurry had a pH of 7.4 and an EC of ∼7 mS cm−1.

2.3.3. Sample Preparation

[26] Upon arrival at the laboratory, all samples were immediately stored in a refrigerator (4°C) and processed within 24 h after sampling. The large groundwater samples were concentrated using a tangential flow filtration system developed for rapid recovery of diverse microbes in drinking water [Hill et al., 2005]. The system was composed of a Masterflex I/P precision brushless pump (model No. 77601–10, Cole-Parmer), Masterflex C-Flex 50-A tubing (Cole-Parmer), an Optiflux 250 filter (Fresenius), and a container for collecting retentate water. Filtrations were conducted at ∼250 mL min−1 with no more than 30 psi pressure. After filtration, 500, 100, 10, 5, and 1 mL subsamples of the retentate water (2000–2500 mL) were subjected to membrane filtration. For soils, 25 g samples were suspended in 100 mL phosphate buffered saline (PBS) in 250 mL centrifuge bottles and homogenized for 15 min on a wrist-action shaker. The samples were gently centrifuged (500 × g for 10 min), supernatants were transferred into a 100 mL bottle, and sediments discarded. The supernatants were serially diluted (100, 10−1, 10−2) and all solutions were used for membrane filtration. For manure slurries, 10, 1, 0.1, 0.01, and 0.001 mL of each sample were suspended in 50 mL PBS, respectively, passed through four-layer gauze to remove larger manure particles (if necessary), and subjected to membrane filtration.

2.3.4. Detection and Enumeration of Enterococcus spp.

[27] Samples were filtered through 50 mm diameter 0.45 μm pore size cellulose nitrate filters (Millipore). Filters were incubated onto mEI agar at 41°C for 24 h. Bacterial colonies with a blue halo were considered to be Enterococcus spp. (U.S. Environmental Protection Agency (USEPA) (2002), Method 1600: Enterococci in water by membrane filtration using membrane-Enterococcus indoxyl-β-d-glucoside agar (mEI), EPA-821-R-02-022; available at

[28] Standard bacteriological protocols recommend cfu density ranges of 25–250 cfu on 100–150 mm diameter plates (U.S. Food and Drug Adminstration, 2001, Aerobic plate count, Chap. 3 in Bacteriological Analyical Manual, available at∼ebam/bam-3.html1#6-international) or 20–60 cfu on 50 mm diameter plates (USEPA, 2002), if practical. Given that the cfu count of Enterococcus spp. per unit volume of contaminated groundwater or soil sample can be very small (unpublished data), we considered that 1–60 cfu on a 50 mm plate was an acceptable count range. For each sample, one plate was selected for enumeration of Enterococcus. Concentrations were expressed as cfu g−1 for soils or cfu per 100 mL for groundwater and manure slurries.

3. Results and Discussion

3.1. Soil Minicolumn Retention of Microspheres

[29] We investigated the importance of soil type, suspending solution properties, and how the characteristics of the artificial microspheres influenced their retention on soil surfaces. An agglomeration of microspheres in leachate was used as a covariate response variable.

[30] The analysis of variance (general linear model, GLM) indicated that all factors (soil, microsphere types, and suspending solution) and their interactions had a significant impact on the retention of microspheres in the soil columns (p < 0.001 for all the three factors and their interactions, overall adjusted r2 = 0.94). Similarly, the same factors and their interactions had a significant impact on the percent agglomeration of microspheres in leachate (p < 0.001 for all factors and their interactions, except the suspending solution factor for which p = 0.023; overall adjusted r2 = 0.88). A direct comparison between the hydrophobic and hydrophilic microspheres indicated that their retention behavior was dissimilar in ddwater and <0.45 micron solution, but more similar in the <1 micron solution, and that soil type had an influence. The use of manure filtrates reduced the variability in retention among the hydrophilic microspheres (Figure 2).

Figure 2.

Retention of the microspheres in the 1-g soil minicolumns as a function of the soil type and type of suspending solution. (A) Open circles and (B) open squares are hydrophobic microspheres while (C) solid circles, (D) solid squares, (E) solid triangles, and (F) solid diamonds are microspheres of variable hydrophilic character (Table 2). Suspending solution: W, distilled sterile water (ddwater), <0.45, manure solution filtered through a 0.45 μm filter, <1, manure solution filtered through a 1 μm filter. Error bars are one standard error (SE) from the mean.

[31] Relative to hydrophilic microspheres, hydrophobic microspheres were retained preferentially in all treatments (Figure 2). Initial assessment indicated that of the three types of suspending liquid, ddwater, resulted in retention of the various microspheres being inversely related to their surface charges. The sandy loam soil particles tended to collapse during wetting, while the aggregates of the other two soils maintained their structure throughout. The silty clay soil tended to expand under wetting, and thus reduce the volume of active pores. Under these conditions soil type affected how the manure components in the suspending solution influenced retention of microspheres. When the <0.45 micron solution was used, retention rates decreased in the sandy loam soil but increased in the silty clay soil columns. The <1 micron solution generally led to increased microsphere retention except in the silt loam soil (Figure 2).

[32] In the silty clay soil, increased dry matter in the suspending solution resulted in a greater retention of microspheres. In the silt loam soil, the only soil that maintained its structure during wetting, the increase in the dry matter content of the solution resulted in decreased retention.

[33] The agglomeration of microspheres in the leachate gave an indirect indication of the role of the suspending solution on the magnitude of hydrophilic and hydrophobic interactions. Comparison between the rates of microsphere agglomeration in leachate with the rates of microsphere retention provided insight into the role of the charge and hydrophobic interactions. Overall, hydrophobic microspheres had greater agglomeration rates in leachate as well as greater retention rates in the soil minicolumns for all suspending solutions (Figure 3) and soils (Figure 4) than did hydrophilic particles. In general, there was an inverse relationship between agglomeration in leachate and retention within columns (Figures 3 and 4). This suggests that both phenomena were related to the hydrophobic character of the microspheres.

Figure 3.

Covariance of the microsphere retention and microsphere agglomeration in soil minicolumns, variation with type of suspending solution. Microsphere types A and B are hydrophobic, while types C–F vary in their hydrophilic characteristics (see Table 2).

Figure 4.

Covariance of the microsphere retention and microsphere agglomeration in soil minicolumns, variation with soil type. Microsphere types A and B are hydrophobic, while types C–F vary in their hydrophilic characteristics (Table 2).

[34] The possible role of each considered factor in the retention of the microspheres was evaluated through PLS regression analysis (Figure 5). When the results for the retention of all hydrophobic and hydrophilic microspheres were combined, the resulting PLS loading plot (Figure 5a) indicated that soil parameters made an important contribution to the variability of microsphere retention. However, this variability was not uniformly correlated with that of the measured soil properties and therefore was not directly explained by them. In contrast, the charge properties of microspheres and manure type (suspending solution) seem to have been the main driving factors in microsphere retention. Retention was inversely proportional to microsphere charge, but increased with the dry matter content of the suspending manure solution (Figure 5a). Variability in the results for the retention of microspheres was not readily explained by their size (Figure 5a), although it appeared to be of some importance for hydrophobic microspheres (Figure 5b).

Figure 5.

Partial least squares analysis of the impact of potential predictors on the retention of artificial microspheres in soil minicolumns and their agglomeration in the leachate regression-loading graphs. The terms in parentheses associated with the PLS components represent x variance, r2. Parameters on the graphs (see Tables 1 and 2): %OM: organic matter content of soil, %sand, %silt, %clay: textural composition of the soils by weight, CEC: cation exchange capacity of the soil, size: diameter of microsphere, charge: surface charge of microspheres, charge dens.: surface charge density of microspheres, manure DM: dry matter content of manure solutions. All parameters have been used for the analyses of (a) “all microspheres” and (c) the hydrophilic microspheres; (b) hydrophobic microspheres analysis did not include charge properties as both had surface charges ∼0 meq g−1.

[35] Separate assessments for the hydrophilic microspheres showed that when all solutions were considered there was no indication that microsphere retention was significantly dependent on their charge or size properties (Figure 5c). Soil characteristics and the dry matter content of the suspending solution explained both the total variance of the observed data and the variance of the means. Retention increased with increased dry matter content in the suspending manure solution (Figure 5c), similar to the results for retention of hydrophobic microspheres (Figure 5b).

[36] Agglomeration in leachate, as defined in our experiment, was inversely correlated with the charge on the microspheres (Figure 5a). It is expected that both suspending solution type and the impact of the soil on the leachate electrochemical parameters would affect agglomeration in the leachate. One notable difference from the retention data was that suspending solution was only a minor contributor to microsphere agglomeration (Figure 5a). However, microsphere charge density and microsphere size had relatively equal but opposite impacts on the agglomeration of the hydrophilic microspheres (Figure 5c). The analysis supported evidence from van Loosdrecht et al. [1987] that increased hydrophobicity led to increased agglomeration of hydrophilic particles (Figure 5c). Overall, variability in the observed data was reasonably well explained by the factors evaluated.

[37] Our results support the finding that manure likely enhances the transport of microbes [Guber et al., 2009] independent of the actual charge of the transported particle. This observation may be somewhat less certain for completely hydrophobic particles (i.e., particles of zero surface charge), but this is an unlikely condition for live microbial cells.

[38] The charges on bacterial cells as estimated by cell microtitration fall somewhere between 0 and 40 meq g−1 dry matter [Noda and Kanemasu, 1984]. The artificial microspheres used for the column retention experiments had a surface charge between 0 and 214 meq g−1 dry matter. In contrast to the known facts for bacterial cells, charges on the surface of microspheres were uniformly distributed. Hence, their response to variations in the charges within the immediate environment should also be more uniform.

[39] As the pH and EC values of the manure filtrates were similar, their other electrical properties were likely similar. Hence, the major differences between the filtrates were the type and amount of suspended solid matter. Given the method of extraction it was expected that the <1 micron solution would have a larger colloidal component than the <0.45 micron solution. Variations in liquid surface tension due to the addition of soluble and insoluble charged components [Reid et al., 1991] had a limited impact on the hydrophobic interactions (Figure 3) as the majority of hydrophobic microspheres were retained on soil, irrespective of the type of suspending solution (Figure 2), with straining a likely contributing mechanism [Bradford et al., 2006]. Obviously, the lack of surface charges precluded any direct interaction with charged loci in the solution or on soil surfaces. Association of the hydrophobic particles with suspended hydrophobic, polar, organic matter from manure [Businelli et al., 1999; Piccolo et al., 1999] would have been expected to increase the hydrophilic interactions, and hence mediate reduced retention. The presence of manure in the suspension was inversely related to the agglomeration of the hydrophobic microspheres (Figure 5b).

[40] Conversely, insoluble soil organic matter would offer more hydrophobic retention sites at the surface of clay-organic complexes. However, the effect of the soil organic matter in the context of the presence of the soluble and insoluble organic compounds from manure is less clear, due to the colinearity of the impact of the soil characteristics.

[41] It is significant that when any microspheres were added to the soil in the presence of manure components, their retention increased, but their agglomeration in leachate was somewhat decreased (Figure 5). This observation is strengthened by the colinearity of the manure and the charge density in the agglomeration-PLS loading graph for the hydrophilic microspheres (Figure 5c), which indicates that the impact of decreasing the surface charge density on a microsphere is very similar to decreasing the manure content, and both are inversely correlated with agglomeration.

[42] Charged microspheres were more likely to remain suspended in an aqueous transporting medium, just as reported for hydrophilic bacteria [van Loosdrecht et al., 1987]. Therefore, retention of microspheres was by and large inversely correlated to their hydrophilic character (Figures 2 and 5a). However, while this was true when comparing hydrophobic and hydrophilic microspheres (Figure 5a), among the hydrophilic microspheres charge properties were important for retention mainly when suspended in water, but in manure solutions the significance of their surface charge properties was less (Figure 2).

[43] Smaller particles have a larger diffusion coefficient, which means that they are more likely to collide and attach to surfaces [Cumbie and McKay, 1999; James and Chrysokopoulos, 2000]. Conversely, larger particles and flocs are more prone to interception and physical straining [Bradford and Torkzaban, 2008]. Although our results for the hydrophobic microspheres support this concept, consistent with Gannon et al. [1991], retention of the hydrophilic microspheres was not influenced by their size within the range investigated (Figure 5b). Retention of bacterial cells is considered to be directly related to the clay content of the soils [Stotzky, 1985]. At finer grain sizes the ratio of colloid size to that of the porous media increases and interception and straining becomes more significant [Bradford et al., 2006]. Any agglomeration of particles will further exacerbate the effect. The analysis of variance confirmed the importance of soil properties in the total variability of microsphere retention (Figure 5).

3.2. Leaching of Microspheres Through Undisturbed Soil Columns

[44] Although the sandy loam soil generated the largest drainage after the first postapplication rainfall simulation, more hydrophobic microspheres leached from the finer textured soils with both sludges (Figure 6). The larger volume of water added with liquid sludge was reflected in greater drainage from the coarse sandy loam soils but not from the finer textured soils. Unc and Goss [2006] previously discussed the mechanisms behind this. They concluded that surface-applied solid organic wastes protect the structural stability of soil pores and thus favor infiltration. This was most obvious for all soils at the second irrigation event. For the subsequent irrigation events the counts of microspheres in drainage from the different treatments were more uniform. For all soils receiving solid sludge the vadose-zone transfer of hydrophobic microspheres was uniform from event to event. For later events the concentration of microspheres declined such that more pore volumes of drainage were required to release a similar total number.

Figure 6.

Impact of sequential discontinuous rain-simulations on undisturbed large soil columns amended with municipal sewage sludge's on the cumulative counts of hydrophobic microspheres in drainage; solid diamonds: clay soil-solid sludge, solid triangle: clay loam soil-solid sludge, solid circle: sandy loam soil-solid sludge, open diamond: clay soil-liquid sludge, open triangle: clay loam soil-liquid sludge, open circle: Sandy loam soil-liquid sludge.

[45] In structured soils, even the transport of hydrophobic bacterial-sized particles occurred as a function of the soil's macropore network with enhanced vadose-zone transport at smaller drainage volumes from finer-textured soils than the coarser sandy soil. This confirmed both the significance of structure over texture for soil's hydraulic behavior in the macropores capable of transporting bacterial-sized particles [Lin et al., 1999] and the role of organic waste in modifying the transport patterns through the vadose zone of any soil type [Unc and Goss, 2006].

3.3. Field Scale Fecal Indicator Transport Into the Unconfined Aquifer

[46] Groundwater levels were constant for the first 2 d and rose by 0.2 m during day 3 and 4 of the irrigation, following application to blocks immediately adjacent to the well on day 2. Subsequently, groundwater levels declined to pre-irrigation levels over 10 d. Resumption of the irrigation after a 2-d break caused an intermittent, smaller (0.1 m) water level rise on day 8.

[47] Breakthrough curves showed a rapid rise in Enterococcus spp. within 24 h of the start of irrigation with a nearly 4 log increase in concentrations (Figure 7). Two-thirds of that increase occurred prior to any changes in water level. A double peak resulted from the break in irrigation between t = 4 d and t = 6 d. Return to near background concentrations (<1 cfu/100 mL) occurred after ∼10 d.

Figure 7.

Breakthrough curves of (a) Enterococcus, chloride, (b) electrical conductivity, and change in groundwater level resulting from border flood irrigation with manure slurry of an agricultural field. Irrigation was temporarily halted between t = 4 d and t = 6 d. Irrigation was complete by t = 10 d.

[48] For chloride and electrical conductivity the double peak in the breakthrough curves was less defined as each had only a minor increase of 5 mg L−1 and 100 mS m−1, respectively, that lasted for the duration of the monitoring period (Figure 7). Peak concentrations were delayed relative to Enterococcus spp. and may have occurred after cessation of the monitoring at 20 d. Onset of electrical conductivity increase coincided with the onset of water level increase on day 3.

[49] Differences in breakthrough curve dynamics between the colloidal cells of Enterococcus spp. and conservative solutes are partly due to a much greater relative difference between background groundwater concentration and irrigation water quality. For Enterocccus spp., the difference between background groundwater concentration and source (manure) concentration was eight orders of magnitude, while the chloride concentration in the slurry was approximately 3.4 times greater than in background groundwater: less than half an order of magnitude.

[50] Salinity and chloride are transported conservatively and may provide alternate evidence for the role of advection by preferential flow through the vadose zone from the land surface. Well samples (taken from a well screen with 4.5-m saturated thickness) are a mix of background groundwater and recharge associated with the irrigation. Assuming a simple linear mixing model, a mixing ratio can be calculated based upon the change in salinity concentration of groundwater relative to background (Table 5). Comparing peak concentrations of chloride and electrical conductivity, a mixing ratio of ∼98:2 of background groundwater to manure slurry recharge would account for the observed increase in groundwater concentration. At this dilution factor, and given known concentrations in the manure slurry, the concentration of Enterococcus spp. in the monitoring well samples would have been on the order of 105 cfu/100 mL, if no attenuation had occurred in the overlying unsaturated zone. Actual peak concentrations were three orders magnitude smaller, of the order of 102 cfu/100 mL.

Table 5. Linear Mixing Model Between Manure Slurry and Groundwatera
Mixing Ratio Groundwater:Manure Slurry (%)Chloride Concentration (mg L−1)Electrical Conductivity (mS m−1)
  • a

    Peak concentration of chloride and electrical conductivity were 79.9 mg L− 1 and 2480 mS m−1, respectively.


3.4. Cross-Scale Comparison of Results

[51] The soils used in the field experiment were predominantly sandy loam in texture and had similar properties to those used for the minicolumn and undisturbed soil column experiments. Enterococcus spp. have hydrophilic surfaces and are similar, albeit slightly smaller in size, to the small hydrophilic microspheres used in the minicolumn tests, but approximately five times smaller than the hydrophilic microspheres applied to the undisturbed soil columns. The liquid manure slurry was unfiltered, similar to the liquid municipal sludge used for the undisturbed soil experiment.

[52] We determined the degree of consistency in colloid retention across the three scales of observation in our experiments according to filtration theory [Bradford et al., 2006; Logan et al., 1995; Sen, 2011; Yao et al., 1971]. Filtration was assumed to be a first order removal process acting in addition to the advection-dispersion-diffusion processes of variably saturated transport [Ginn, 2002; Sen, 2011]. The fractional penetration (total nonretained fraction), fp, of colloids to a distance, X, from the source is related to the filtration coefficient, λ, such that:

display math

For steady state transport experiments, fp is equal to the relative peak concentration c/c0, where c0 is the inflow concentration and c is the (steady) outflow concentration. The filtration coefficient, λ, is a function of buoyancy or gravity, diffusion, interception, and surface-surface interaction forces between the colloid and grain surfaces. The Happel-sphere-in-cell model has commonly been used to estimate λ [Logan et al., 1995; Rajagopalan and Tien, 1976]:

display math

where θ is the volumetric soil water content, dc is the effective grain size diameter, α is the collision efficiency, and η is the collector efficiency. Water content, grain size diameter, and collector efficiency can be determined based on the soil and colloid physical properties (ibid.). The collision efficiency, α, is a fitting parameter that accounts for unfavorable chemical, electrostatic, and other interactions between the colloid and surfaces that are not accounted for by η. Colloid filtration theory applies to the attachment of charged colloids in saturated porous media [Yao et al., 1971]. However, in variably saturated media additional processes occur, such as variable air-water interface attachments [Schäfer, 1998a, 1998b] and variability in straining and effective pore diameter [Bradford and Torkzaban, 2008], that are not explicitly accounted for by colloid filtration theory. Here, α intrinsically accounts for the additional effects of variably saturated flow [e.g., Sen, 2011; Torkzaban et al., 2008] and transient flow conditions not accounted for in (1) and (2). The collision efficiency α is typically obtained by computing η and by empirically determining relative concentration peaks, c/c0, fractional penetration, fp, by curve-fitting to breakthrough curves of colloids [e.g., Harter et al., 2000; Torkzaban et al., 2008], or directly from the slope of sorbed or attached colloid profiles [e.g., Harter et al., 2000]. Using (2) in (1) and solving for α, gives [Abudalo et al., 2010]:

display math

We determined values for α (Table 6) by solving (2) with experiment-specific values of η, and λ was estimated using fp from the minicolumn experiments and from the undisturbed soil column experiments, while c/c0 was obtained from the field irrigation breakthrough curve (Figure 7). For the field experiment, c/c0 = 0.001 was obtained by considering the actually observed peak concentration (102 cfu/100 mL) and the theoretical, unattenuated concentration (105 cfu/100 mL) estimated by applying the observed mixing ratio of the conservative tracers (chloride, EC, Table 5) to the measured manure concentration. In the minicolumn experiments the collision efficiencies for comparably sized microspheres (“D” and “F”) in sandy loam soil after application of <1 micron solution manure slurry ranged from 0.1 to 0.2, and for the undisturbed soil column experiments (sandy soil, liquid slurry), α ranged from 0.05 to 0.06.

Table 6. Computed Collision Efficiencies, α, for the Minicolumn, Undisturbed Soil, and Field Irrigation Experimentsa
Experimentθ (%)u* (m s−1)c/c0α

    C, clay; CL, clay-loam; SL, sandy-loam; LB, liquid treated municipal sewage; DW, dewatered municipal sewage sludge (biosolids).

  • a

    Unless noted otherwise, we assume that dc = 1.25 × 10−4 m, μ = 9.61 × 10−4 kg m−1 s, ρρ = 1050 kg m−3, and ρf = 997.77 kg m−3.

L = 0.0364 m and dp = 1.5 × 10−6 m
fp = 50%0.351.00 × 10−45.00 × 10−11.24 × 10−1
fp = 20%0.351.00 × 10−42.00 × 10−12.89 × 10−1
fp = 10%0.351.00 × 10−41.00 × 10−14.13 × 10−1
fp = 5%0.351.00 × 10−45.00 × 10−25.37 × 10−1
Undisturbed Soil Columns
L = 0.5 m and dp = 3.7 × 10−6 m
C-DW-10.372.03 × 10−59.16 × 10−76.24 × 10−2
C-DW-20.399.53 × 10−53.98 × 10−81.53 × 10−1
C-DW-30.405.61 × 10−52.39 × 10−71.22 × 10−1
C-DW-40.392.28 × 10−43.98 × 10−81.89 × 10−1
C-LB-10.341.29 × 10−41.63 × 10−45.89 × 10−2
C-LB-20.398.85 × 10−51.12 × 10−61.18 × 10−1
C-LB-30.415.79 × 10−52.39 × 10−71.38 × 10−1
C-LB-40.402.63 × 10−48.48 × 10−92.21 × 10−1
CL-DW-10.303.29 × 10−51.20 × 10−75.48 × 10−2
CL-DW-20.279.45 × 10−54.24 × 10−86.05 × 10−2
CL-DW-30.371.01 × 10−44.24 × 10−81.37 × 10−1
CL-DW-40.322.44 × 10−43.98 × 10−81.11 × 10−1
CL-LB-10.277.27 × 10−52.24 × 10−42.62 × 10−2
CL-LB-20.286.10 × 10−57.97 × 10−85.73 × 10−2
CL-LB-30.337.58 × 10−51.99 × 10−78.17 × 10−2
CL-LB-40.322.38 × 10−43.98 × 10−81.06 × 10−2
SL-DW-10.281.21 × 10−41.20 × 10−76.26 × 10−2
SL-DW-20.271.03 × 10−43.98 × 10−86.11 × 10−2
SL-DW-30.279.73 × 10−53.98 × 10−86.10 × 10−2
SL-DW-40.292.01 × 10−43.98 × 10−88.19 × 10−2
SL-LB-10.251.94 × 10−46.37 × 10−74.90 × 10−2
SL-LB-20.267.82 × 10−57.97 × 10−84.66 × 10−2
SL-LB-30.261.06 × 10−41.59 × 10−75.01 × 10−2
SL-LB-40.262.83 × 10−41.99 × 10−75.90 × 10−2
Field-Scale Irrigation
L= 4.5 m, μ = 9.40 × 10−4 kg m−1 s, ρρ = 1011 kg m−3, ρf = 999 kg m−3, and dp = 8.4 × 10−7 m
Enterococcus0.351.04 × 10−41.00 × 10−39.41 × 10−3

[53] In the field, for median-sized Enterococcus spp., (0.84 μm) having a density of 1.11 g cm−3, with a soil volumetric water content of 35%, an effective media grain size of 125 μm, a travel distance of 4.5 m to the monitoring well (through the unsaturated zone to the water table), and a mean travel time of 12 h, field-scale collision efficiency (α) was 9.4 × 10−3. If travel time was increased to 48 h, the value of α was reduced to 4.0 × 10−3. In drier soil with an effective volumetric water content of 25% the fitted value of α was also smaller, 5.7 × 10−3. For the largest cells of Enterococcus spp. (diameter = 1.25 μm) with the same porous medium, sampling geometry, and a volumetric water content of 35%, the value of α was 1.01 × 10−2 for a 12 h travel time. For the slower travel time it reduced to 4.6×10−3 and was the same value (5.7 × 10−3) as that for the smaller cells in the drier soil.

[54] The results also showed that the collision efficiency for variably saturated, transient flow decreased by half an order of magnitude for every order of magnitude increase in total transport distance (Figure 8). For comparison, we computed α for the experiments by Torkzaban et al. [2008], who used microspheres similar to “D” of our minicolumn experiment in variably saturated but steady state flow experiments using 10 cm columns filled with Ottawa quartz sand (“MIX”), with a mean gain size of 0.24 mm and a uniformity coefficient of 3.06 in a high ionic strength solution (60 mM). The ionic strength in their experiments was nearly an order of magnitude smaller than the ionic strength in the manure slurry, but similar to the ionic strength of the (diluted) groundwater in the field irrigation experiment with manure: ∼120 mM. Despite these differences, the resulting effective collision efficiency (α = 0.07) was indeed intermediate between our 3.6 cm minicolumn and 50 cm undisturbed soil column experiment, and consistent with the observed length-scale dependent trend of the collision efficiency, α (Figure 8), albeit slightly smaller. This value was consistent with the decreases in retention observed for hydrophilic microsphere transport in sandy loam soils at smaller ionic strength (Figure 2), and hence, smaller observed collision efficiencies. However, travel distance or transport scale appeared to be an overriding factor in determining the magnitude of the collision efficiency when applying the classic filtration concept to these data.

Figure 8.

Length-scale dependence of the empirically determined collision efficiency (Rajagopalan-Tien model) in four variably saturated experiments and in the saturated column experiments with E. coli strain UCF94 by Lutterodt et al. [2011]. Data points at 0.1 m are computed from the “MIX” sand experiments “Hi100” and “Hi80” by Torkzaban et al. [2008]. Data points for Lutterodt et al. [2011] are computed from original experimental data (courtesy of G. Lutterodt, unpublished data, 2011). The limiting case (thin straight line) represents a constant, distance-independent fractional penetration of 0.04%. The generalized power function (5) was computed for A = 0.04, β = 0.43, and γ = 0.93.

[55] Lutterodt et al. [2011] demonstrated the scale dependence of the filtration coefficient in a series of experiments with two strains of Escherichia coli in a relatively long 25 m saturated quartz sand column with four sampling ports at 6.25 m, 12.5 m, 18.75 m, and 24.95 m distance from the column inlet. For the two different E. coli strains (UCF94 and UCF131), which demonstrated low levels of attachment, the α values based on the total mass injected and the total mass under the breakthrough curve at each of the four ports decrease significantly with transport distance (Figure 8). The collision efficiency for UCF94 was greater than for UCF131. On a log-log scale, collision efficiencies for the two strains decreased at the same (power law) rate with distance.

[56] Bradford et al. [2006] and Torkzaban et al. [2008] argued that the depth-dependent retention of hydrophilic colloids (unfavorable to attachment) was mostly due to a weak association of colloids on solids surfaces as a result of straining and lateral movement of colloids along grain surfaces into stagnant flow zones at grain-grain contacts within the pore system. Bradford and Bettahar [2005] suggested that straining, physical filtering, or Langmuirian deposition and filling of pores would lead to relatively larger deposition immediately below the colloid source than at some distance from the source, leading to a distance-dependent removal (filtration) coefficient. Bradford et al. [2003] and Bradford and Bettahar [2005] therefore added a depth-dependent filtration term to the classic filtration problem, which decreases as a power function of depth. With respect to α, their straining filtration model can be rewritten:

display math

where α(X) is a distance-dependent collision efficiency and α0 is the effective collision efficiency at the scale of the porous media grain diameter. Experimentally, Bradford et al. [2003] determined that β is 0.432, thus, in fact, confirming that the overall attenuation rate typically decreases at a rate of approximately half an order of magnitude per order of magnitude increase in transport distance (Figure 8). However, their evidence has been restricted to relatively short transport scales (on the order of 0.1 m). Our data suggest that the conceptual framework (4) can be extended to significantly larger length scales. The combined data from our three experiments, from the unsaturated experiments by Torkzaban et al. [2008], and from the large distance experiments specifically of UCF94 by Lutterodt et al. [2011] do follow a power law (Figure 8). However, the power law slope appears to increase from ∼0.4 at short distances to a β value near unity at large distances. By taking the logarithms on both sides of (3), it can be shown that β = 1 represents the limiting case (largest slope), i.e., when fp reaches a constant value with depth (no further filtration beyond some finite distance). The apparent change in the slope in Figure 8 suggests the following general power law equation for distance-dependent collision efficiency, α(X):

display math

where A is a scaling coefficient, β represents the power law slope at short distances, and γ represents the power law slope for large travel distances (0 ≤ β < γ ≤ 1). The error-function (erf[X]) is here used to transition the power exponent between these two values as a function of distance X, which we suggest may be further scaled within the error-function, e.g., erf([DX]E), where D and E are scaling coefficients. Using β = 0.43, as suggested by Bradford et al. [2003], and γ = 0.93, as suggested by the power law scaling of local collision efficiencies by Lutterodt et al. [2009, 2011] in (5), provides a close fit to the empirically determined collision efficiencies of all experiments (Figure 8).

[57] We point out that such depth-dependent collision efficiency is not a test of the specific straining processes discussed, e.g., in the work of Torkzaban et al. [2008], but rather that one or, more likely, several mechanisms control retention or filtration in a strongly scale-dependent fashion. Besides straining, these processes likely include porous media heterogeneity, variability, and selective filtration among colloids, preferential flow paths, and depth-dependent presence of organic matter. Attachment to solid-water or air-water interfaces can also be confounded with straining [Gargiulo et al., 2008]. Lutterodt et al. [2009, 2011] suggested that distance-dependent collision efficiencies of E. coli are due to inter- and intrastrain variability in surface properties including variable zeta-potential, motility, hydrophobicity, and outer-membrane protein composition.

[58] Retention in stagnant zones and at grain-grain contacts are consistent with the concept that unretained colloids, including bacteria, experience a spectrum of pore sizes and pore velocities, but with a bias toward larger values of both and, hence, more rapid transport [Bradford et al., 2006; Ginn, 2002; Harter et al., 2000; Unc and Goss, 2003]. The breakthrough of Enterococcus spp. under irrigation in the field experiment was faster than that of solute tracers as a result of either size exclusion or preferential transport of colloids through macropores in the unsaturated zone. Since we did not directly measure Enterococcus spp. concentration and attachment in the vadose zone, it is not possible to determine directly whether the bacteria captured during an irrigation event originated from a prior event and have been remobilized [Harvey et al., 1993] as a result of hydrodynamic interactions and modified attachment and collision efficiencies induced by mixing of manure into groundwater [Cail and Hochella, 2005; Schinner et al., 2010] or whether they are due to rapid breakthrough from the current irrigation event. However, we argue heuristically that consistency between our laboratory and field results, and with rapid transport observations reported in the literature together with the breakthrough of chloride and salts immediately after the irrigation are all indicative of rapid transport (not remobilization) of Enterococcus spp. through the vadose zone being the most plausible explanation, consistent with (5). Results from the large core tests suggest that delayed remobilization of bacteria-sized particles (>1 month since the last manure application) can occur and hence could also happen in the case of Enterococcus spp., but likely at concentrations that are significantly smaller than those associated with the irrigation concurrent with the land application of wastes.

4. Conclusions

[59] Small-scale tests show that the hydrophobic characteristics of microspheres govern their initial retention to soil surfaces independent of the properties of the soil or the transporting solution. Differences in individual charge and size properties of hydrophilic microspheres, which are more representative of bacterial colloids, are only significant for the retention in low-ionic strength water with minimal or no dissolved organic carbon, but are of no consequence where manure extracts are present. The inverse relationship between microsphere agglomeration in the collected leachate and their retention in the soil suggests that the interaction between manure components and soils governs the retention of microspheres independent of other properties of the same microspheres. Significant contaminant transport through undisturbed, unsaturated soils occurred at drainage volumes equivalent to <1% PV for finer-textured soils and <6% PV for coarser sandy-loam soils. While initial contaminant movement was enhanced by the application in liquid wastes, consistent with observations in the small-scale tests, long-term vertical transport of bacterial-sized particles and Enterococcus spp. (an indicator for pathogenic organisms) was similar for different soil and waste types. Using classic filtration theory for an integrated analysis across multiple scales and multiple experiments, including relevant experiments reported in the literature, indicates that the role of filtration forces (buoyancy, diffusion, inertial forces, surface-surface interactions) diminishes significantly with transport distance as indicated by the significant, solution and colloid-independent, consistent decrease in collision efficiency as the experimental scale increased. Our results suggest that a modified filtration theory with distance-dependent collision efficiency, while ignoring the known biological, chemical, and physical complexity of colloid transport processes in an organic waste treated soil, may provide a first order approximation adequate to some relevant regulatory and water management problems.


[60] Funding was provided by the New Mexico State University Agricultural Experiment Station, by the United States National Institute for Food and Agriculture grant 2007-02855 and by the Resources Management and Environment Program of the Enhanced Partnership between the University of Guelph and the Ontario Ministry of Agriculture and Food. We thank Joanna Passmore for technical support in studies with undisturbed soil columns.