Microbial and solute transport through intact vadose zone cores of heterogeneous alluvial gravel under variably saturated conditions

The movement of bacterial and viral pathogens through soil and vadose zone and subsequently into groundwater is a major public health concern. There are relatively few studies on the transport and fate of microbes through variably saturated vadose zone media compared with their transport in the soil and saturated groundwater zones. In this study, we investigated the transport of Escherichia coli, F‐RNA bacteriophage MS2, and a conservative solute tracer bromide through three intact vadose zone cores, under saturated (discharge rate ∼100 mm h−1) and unsaturated (discharge rate 10 and 0.5 mm h−1) flow conditions. The vadose zone media were sandy gravel overlying a sand lens in core 1, a heterogeneous SG mix in core 2, and SG with an open framework gravel lens through the middle of the core in core 3. The three flow regimes resulted in different transport characteristics through each of the cores. As expected, microbial transport through all cores was higher under saturated conditions, compared with unsaturated conditions. Overall, E. coli removal was consistently greater than that of MS2 phage irrespective of core media or flow conditions. There were relatively minor removals (factors of 1–2.5) of both microbes under saturated conditions, reductions of 2–3 orders of magnitude under the high flow unsaturated conditions, and almost complete removal (4 to >5 orders of magnitude) under the low flow unsaturated conditions. The much greater removal of microbes under unsaturated conditions has significant implications and potential benefits for land management decisions.


INTRODUCTION
Groundwater comprises the vast majority (∼ 95%) of the world's usable fresh liquid water and is an important source outbreaks associated with groundwater (Jin & Flury, 2002). Bradford and Harvey (2017), in their review of research needs involving pathogens in groundwater, stated that there are considerable knowledge gaps regarding the transport and fate of microbes through the heterogeneous subsurface to groundwater (i.e., variably saturated media in the vadose zone).
Contamination of groundwater by pathogenic microbes can be generated from a range of activities, such as effluent applications to land, leaking sewers (Gotkowitz et al., 2016), grazing of pasture animals (Moriarty et al., 2008), and on-site wastewater treatment systems, and many of these microbes are known to survive a considerable time in the environment (Engstrom et al., 2015;Pedley et al., 2006;Schijven et al., 2017). Alluvial gravel aquifers are very productive and economically important, providing irrigation supply and potable water for farm households and nearby communities. However, the overlying soils and vadose zone materials are often thin and stony meaning that these aquifers are vulnerable to microbial contamination. The risk of microbial transport through the subsurface from farming practices often depends on whether the soil surface and vadose zone becomes saturated (Clemens et al., 2020;Close et al., 2008Close et al., , 2010. For on-site wastewater treatment systems, saturated conditions are more common and removal of microbes depends more on the soakage trench media and the properties of the underlying vadose zone material. In this paper the vadose zone is defined as the zone from the bottom of the soil layer to the groundwater table. Microbes need to be transported through the soil and underlying vadose zone before entering the groundwater system. Many studies have been carried out regarding microbial transport through soils (Aislabie et al., 2001;Jamieson et al., 2002;Jiang et al., 2008;Karathanasis et al., 2006;McLeod et al., 2004;Pang et al., 2008;Roodsari et al., 2005;Sephernia et al., 2014;Wang et al., 2013Wang et al., , 2014. However, far less is known about transport of microbes through the underlying vadose zone due to its lack of accessibility (Close et al., 2010;Weaver et al., 2016). This is also the case in comparison with saturated groundwater systems. Pang (2009) collated microbial removal rates from field experiments and large lysimeter studies and found that far fewer studies have been undertaken on microbial removal in vadose zones compared with microbial removal in either soils or groundwater. The vadose zone is characterized by much lower levels of organic carbon and naturally occurring microbes compared with soils and may vary in thickness from virtually zero (high water table) to tens or hundreds of meters where the groundwater table is deep. Porosity and permeability can vary widely depending on the vadose zone media.
Alluvial gravel systems, which comprise structured and heterogeneous layers of sandy gravel (SG), sand (S) and open framework gravel (OFG) lenses, have been identified

Core Ideas
• Removal of Escherichia coli and MS2 phage in vadose zone alluvial gravel examined under variable saturation conditions. • E. coli showed greater removal compared with MS2 phage due to larger size. • Both microbes showed minor removal (factors of 1-2.5) under saturated conditions. • Both microbes showed significant removal under high (2-3 logs) and low (4-5 logs) flow unsaturated flow. • Land management avoiding saturated conditions will greatly reduce microbial transport.
as high-risk systems of microbial transport, particularly as water saturation increases. In a probable worst-case situation, Sinton et al. (2005) introduced tracer microbes to an effluent flood irrigation system in Canterbury, New Zealand, where there was a 16.8 m deep vadose zone comprising alluvial gravel media. The tracer bacteria were observed in a well approximately 100 m down gradient of injection within one day, implying vertical effluent transport velocities through the vadose zone of at least 360 m d −1 . Sinton et al. (1980) observed tracer bacteria in wells as far away as 1 km from the injection source in gravel aquifers in New Zealand. Transport velocities in alluvial gravels aquifers can be in the order of 10-100 m d −1 Sarris et al., 2018), with the flow direction at a local scale being influenced by the heterogeneous nature of the substrates and preferential flow paths including those formed by networks of linked OFG. Water transport through the vadose zone is highly affected by the degree of water saturation, being very slow under low saturation conditions and increasing exponentially as the water content approaches saturation (Simunek et al., 2003;van Genuchten et al., 1991). As microbes are particles, their transport is even more influenced by water saturation status as only the small pores will be transmitting water and contaminants at low water contents, and hence microbes will be removed more efficiently through various filtration processes within the smaller pores. The factors influencing the transport and survival of Escherichia coli through the unsaturated zone has recently been reviewed by Engstrom et al. (2015). They indicated that the key factors include: (1) E. coli properties (surface charge, hydrophilicity, growth state, presence of surface polymers, and motility); (2) porous media properties (grain size, size distribution, clay content, surface area and charge, physical and chemical heterogeneities); (3) solution characteristics (ionic strength, pH, and content of nutrients, surfactants, and dissolved organic matter); and (4) system characteristics such as temperature, saturation, infiltration rate, land use, presence of worm burrows, vegetation and biofilm, solution chemistry, and moisture content (Engstrom et al., 2015). These factors have been derived predominantly from studies in unsaturated soils and there remains a lack of studies looking at unsaturated media below the soil zone.
Agricultural intensification is occurring worldwide due to increasing demand for food and agricultural products. When high levels of irrigation occur (especially when applied as flood irrigation) or when more moderate volume irrigation occurs close to a natural rainfall event, it can cause increased percolation into the soil which can readily approach or exceed field capacity . At field capacity, the vertical flow through the soil column towards the vadose zone is often near saturated flow conditions. Preferential flow through cracks and macropores can be an important process for saturated flow conditions as water transport is rapid through these large macropores and there is little filtration of microbes taking place in this situation (Bradford & Harvey, 2017;Close et al., 2008). Steady-state saturated flow in the vadose zone is rare in nature; saturated flow predominantly occurs as a transient phenomenon. The wetting front (saturated flow) generated by increased water percolation into the column (potentially by irrigation) generally moves downwards into unsaturated areas of the vadose zone. This increased percolation may increase the downward movement of pathogenic microbes towards the vadose zone. Clemens et al. (2020) investigated the attenuation and transport of rotavirus, MS2 bacteriophage and DNA-labeled glycoprotein-coated silica nanoparticles (DGSnp) in 40 cm intact cores of silt loam over gravels dosed with wastewater under saturated conditions. They found that that, in comparison with rotavirus and DGSnp, MS2 recovery through the 40 cm core was significantly greater with the average mass recovery relative to KBr of 7 and 1%, respectively. The use of intact cores rather than repacked columns is recommended for microbial transport studies so that the natural structure is preserved and macropore and micropore flow remain viable, as opposed to repacked cores where the pore connectivity is disrupted and cannot be replicated McLeod et al., 2014). However, the challenge of obtaining such intact cores for alluvial gravel vadose media means that such studies are rare and the microbial transport data for alluvial gravel media are very scarce.
In this study, we determined the transport of fecal indicators (E. coli, bacteriophage MS2) and a solute water tracer bromide (Br) through typical gravel vadose zone materials under saturated flow and unsaturated (low and high) flow conditions. Intact cores of SG material containing representative horizons including OFG and a sand lens (SL) were characterized and used for the transport experiments. Such information is required for the improved estimation of risks associated with microbial loading and contamination under a range of conditions experienced under farming and effluent management situations as well as for the development of accurate models predicting where and when the contamination of groundwater is likely from pathogens found in effluents applied to soil.

Intact core collection and description
Three intact cores of undisturbed heterogeneous alluvial gravel vadose zone material were collected at two sites south of Christchurch, Canterbury, New Zealand. One core was collected from a gravel pit near Burnham (43˚37″ S, 172˚19″ E) and two additional cores were collected near Lincoln (43˚37′19″ S, 172˚28′10″ E). Dann et al. (2009) gave a full description of the Canterbury alluvial gravel vadose zone material and Table 1 outlines the characteristics of the different Facies found in the Canterbury Plains aquifers. Briefly, the alluvial gravels are highly heterogeneous and consist of three major textural types which comprise various SG mixtures (∼90%), SLs (∼5%), and OFG lenses, often with associated secondary fine silts on the upper and lower edges (∼5%). The Burnham site is characterized by shallow soils underlain by a general matrix of SGs interspersed with lenses of highly permeable gravel. The core collected at Burnham was comprised mainly of SG material. The soil at the Lincoln site mostly consists of rounded gravels (0.5-3 cm up to 7-10 cm in diameter) and sandy material with some silt. The cores collected at Lincoln were comprised mainly of SG material with one core containing an OFG lens and another containing a SL. All cores were collected from depths between 1.5 and 3 m (overlying silt loam soils ∼0.5-1 m deep).
To extract the cores, the vadose material was hand dug using extreme care to leave an intact columnar pedestal of vadose zone material (Figure 1). A PVC casing (300 mm internal diameter, 400 mm height) was then let down over the intact vadose zone cores with a gap between the casing and the intact core of generally about 2-3 cm. Highly viscous warm (∼50˚C) petroleum jelly was injected into this annular gap, rapidly cooling on contact with the soil producing a watertight seal and preventing preferential flow down the perimeter of the cores (Cameron et al., 1990). The base of the cores was separated from the ground by pressing a solid steel circular cutting plate under the cores using a hydraulic jack. The cores were transported to the groundwater research laboratory at Institute of Environmental Science and Research Ltd. for hydraulic conductivity and tracer breakthrough tests under saturated and unsaturated conditions. T A B L E 1 Characteristics of the Canterbury Plains gravel aquifers (Dann et al., 2009)

X-ray computed tomography
The use of X-ray computed tomography (CT) has proven valuable for the analysis of soil structure and soil hydro-physical properties in recent years (Clemens et al., 2020;Lamandé et al., 2013;Luo et al., 2010;Naveed et al., 2015;Taina et al., 2008). The intact cores were scanned using X-ray CT to reveal their structure and macropore characteristics. This process involved scanning each core to produce a series of spinal (vertical) and coronal (horizontal) images, producing one image at every 0.5-0.6 mm (700-768 images per core).

Experimental procedures
2.3.1 Tracer and microorganism characteristics, preparation, and analysis Potassium bromide (KBr) was used as a conservative tracer in each core experiment and was assayed using a Br specific ion electrode (Thermo Fisher Scientific) attached to an Orion250A pH/ISE meter.
The tracer bacterial indicator used in this study was E. coli J6-2. It has rod-shaped cells that are 1.0-1.5 μm diameter by 3.0-5.0 μm long, multiple flagella, is motile and is a lactose negative and nalidixic acid-resistant derivative of E. coli K-12 (Bergstrom, 2000;Sinton, 1980). Cells were cultured at 37˚C in brain heart infusion broth (BD BBL), washed twice by centrifugation and resuspended in saline solution, and stored at 4˚C prior to injection within 24 h. Analysis of samples recovered from the core experiments for E. coli J6-2 was by membrane filtration through 0.45 μm pore size filters (Millipore EZ-Pak) placed onto MacKonkey agar supplemented with nalidixic acid. Plates were incubated for 22 ± 4 h at 37 ± 1˚C. Counts were expressed as colony forming units (cfu) mL −1 . The detection limit for E. coli was 0.1 cfu mL −1 .
The viral tracer used in this study was the F-RNA bacteriophage MS2, which is commonly used as a model virus. MS2 phage is an icosahedral phage, approximately 26-27 nm in diameter ). The host strain used for the experiments was E. coli HS(pFamp)R (Debartolomeis & Cabelli, 1991). MS2 phage injection mixtures were prepared by growing plaques to confluence on an overlay assay with tryptone yeast-extract glucose agar. The MS2 phage were scraped in the overlay agar and then separated by shaking in 100 mL sterile water to release the phage. The agar and cellular debris were then removed by centrifugation (200×g for 25 min). The supernatant was then filtered (0.22 μm pore size) and stored at −20˚C until use (Weaver et al., 2013). Assays of water samples from the core experiments was by overlay pour plating of 1 mL volumes of serial dilutions (APHA, 2004). The base and overlay were prepared from tryptone glucose agar, as described in Debartolomeis & Cabelli (1991). Plates were incubated at 35 ± 1˚C for 16 ± 4 h. Counts were expressed as plaque forming units (pfu) mL −1 . The detection limit was 0.1 pfu mL −1 .
F-RNA bacteriophages are known to be very resistant to inactivation (Nasser et al., 1993). The inactivation pattern was assumed to follow the first order die-off model known as Chick's Law (Equation 1) from Maier et al. (2000).
where N 0 is the number of live microorganisms at t = 0, N t is the number of live microorganisms at after time t, t is time, and k d is the inactivation rate coefficient.
Owing to the long injection phase, especially for the low flow rates, inactivation/die-off of E. coli and MS2 phage was tested in the injection mix to check their viability at the end of the injection phase. Samples of the injection mix were kept in the dark, on a rotary mixer to keep the microbial tracers mixed. Samples were taken intermittently for 21 days and analyzed for E. coli J6-2 and MS2 phage. Sizes and zeta potential variations with pH had previously been measured for the E. coli J6-2 and MS2 phage used in these experiments by Pang et al. (2009).

Hydraulic conductivity experiments
The hydraulic conductivity (K) of the cores were measured under saturated conditions to determine the saturated hydraulic conductivity (K sat ). A collection cup and attachment hose were connected to the core's outflow hose at the bottom of the core. Attached to the cup below the outflow hose was a tube to measure hydraulic head. Each core was slowly filled via a hose from the bottom to prevent air bubbles from being trapped within it. Once the core was saturated, the hose was transferred to the top of the core and a constant pressure head was set. A head difference of 5 or 15 cm between inlet and outlet was attained by moving the collection cup either up or down. Once the target head was attained, the amount of water leaving the collection cup via the "collection tube" was weighed after a certain time. The experiments were run for a minimum of 2 h to gain sufficient data. Once the experiment was complete, a piece of geotextile fabric was placed over the core, and it was covered with sand to prevent degradation of the core. The K sat values were calculated using Darcy's equation: where Q is the flow rate, A is the area of the intact core, and I is the hydraulic gradient.
Unsaturated hydraulic conductivities were measured in triplicate for core 2 (SG) at the two unsaturated flow rates using a similar equation with the hydraulic gradient being measured as the head or tension difference measured by the tensiometers located at the top of the column and the tension in the base chamber.

Unsaturated and saturated transport experiments
Three sets of experiments were run on each core to assess the transport of solute and microbes through the materials under saturated, high, and low flow unsaturated conditions. These experiments were conducted on the intact cores using an automated irrigation and base soil tension and leachate collection system ( Figure 2).
The water and tracers were supplied to the top of the core via a logger-controlled solenoid valve, through three equally spaced fine sprayers with anti-drip devices attached. Prior to conducting the experiments, excess petroleum jelly and disturbed material was removed from the top and bottom of the core. The top of the core was covered with fine gravel (2-5 mm in diameter) to evenly disperse the irrigated water over the surface of the core (pretests indicated that the fine gravels did not attenuate microbes). A thin layer of fine gravel was also added to the base of the core to enable good contact with the perforated conical base plate. The core base plate was then connected to a vacuum base tank within which leachate was collected in a 50 mL sample cup, with excess water overflowing into the base tank. The base tank also controlled the applied suction. The base tank was continually evacuated using a vacuum pump to a negative pressure of between −30 and −50 mm water, which was controlled by an attached bubble tower.
Four tensiometers and a temperature probe were embedded into the top fine gravel layer packed with a mixture of sand and glass beads to give good contact, to provide readings of both top tension and temperature. The temperature probe also measured ambient temperature. Pressure changes were measured by differential pressure transducers attached to the tensiometers and the bubble tower, which functioned as a second regulator of the suction. The volume of solution leached from the core was logged using a load cell below the base tank. When water within the base tank reached a designated volume, the tank was pumped out using a small submersible water pump. The sampling frequency varied with the experiment flow rate and collected between 30 and 40 samples per breakthrough curve (BTC). Solution samples were extracted at set times from the collection cup manually during working hours and via an auto sampler (ISCO 3700) during non-working hours. These samples were divided into Vadose Zone Journal F I G U R E 2 Schematic representation of the experimental set up for the unsaturated flow experiments two subsamples, one for Br analysis and one refrigerated for analysis of E. coli and MS2 phage within 24 h. The microbial samples were analyzed in triplicate using at least three levels of serial dilutions.
In the first set of tracer experiments, each core was run under saturated flow conditions with a constant driving head of 25 mm (discharge rate ∼100 mm h −1 ). Deaerated water was used to saturate the core from the bottom up. After drainage had stabilized, the water flow was stopped, the water level was allowed to drop to the gravel surface before irrigating the top of the core with 3 L of tracer solution (to reduce dilution of tracers) at a rate (around 120 mL min −1 ) that kept the hydraulic head at 25 mm. The concentration of tracers in the injection solution for the saturated experiments are shown in Table 2. After the tracer had entered the gravel material at the top of the core, deaerated water was again applied to the surface of the core until the experiment was complete. Leachate was sampled from the core periodically. Core 1 was initially run with just Br and E. coli, with MS2 phage being added to the saturated experiments for cores 2 and 3, and all the unsaturated experiments. Core 1 was rerun for the saturated flow conditions at the end of the experimental period with all three tracers. However, it was noted that there were significant numbers of other bacteria in the column for this last experiment (as seen on the culture plates) indicating that some biofilm growth had occurred in the column. It was felt that disinfecting the column would cause more change in the column transport characteristics than caused by the biofilm growth. Some differences in transport are noted for this experiment and are discussed in the results section. The second set of experiments were run under high flow (discharge or Darcy velocity ∼10 mm h −1 ) unsaturated conditions. These experiments simulated situations of heavy rainfall or irrigation and discharge of septic tank effluent, for which vadose zone media are often unsaturated but close to saturated flow conditions. Irrigation of the cores was applied in short pulses every 5 min to provide a discharge velocity of ∼10 mm h −1 . The cores were irrigated for 3 days prior to injection to establish steady flow conditions. At the time of tracer injection, the source water was changed to a reservoir containing the tracers, Br (50 mg L −1 ), E. coli (1 × 10 5 cfu mL −1 ), and MS2 phage (1 × 10 5 pfu mL −1 ) from which it was pumped into the core via the sprayer system (experimental conditions given in Table 3). After the tracer was injected, irrigation was continued at a discharge velocity of ∼10 mm h −1 .
The third set of experiments were run under low flow (discharge velocity ∼0.5 mm h −1 ) unsaturated conditions. Irrigation of the cores was applied in short pulses every 30 min to provide a discharge velocity of ∼0.5 mm h −1 . The cores were irrigated at this rate for 3 days prior to injection of the tracers. Tracers Br, E. coli and MS2 phage, at the same concentration as used in the second set of experiments, were injected at the same rate as the irrigated water (experimental conditions given in Table 3). After the tracer was injected, irrigation was continued at a discharge velocity of 0.5 mm h −1 .

2.6
Data analysis and modeling 2.6.1 Microbial attenuation The removal of E. coli and MS2 phage was evaluated using peak breakthrough attenuation, the spatial removal rate, and normalized mass recovery. The peak breakthrough attenuation is the log-reduction of the peak effluent concentration C max relative to the input concentration C 0 , log 10 (C max /C 0 ). The spatial removal rate was estimated from the peak breakthrough attenuation over its transport distance, f = log(C max /C 0 )/x, where x is the transport distance. The normalized mass recovery (RB) was calculated from the mass recovery of E. coli or MS2 over that of the Br, RB = microbial mass recovery/Br mass recovery. Abbreviations: RBr, normalized mass recovery relative to Br; f, spatial removal rate f = log (C max /C 0 )/x. a Complete removal of tracer. b Only a single detection of E. coli was observed in this experiment.

Modeling
Transport and attenuation of microbes through the intact cores under saturated and unsaturated conditions was described using an advection-dispersion equilibrium model that incorporates a first-order removal term: where C is the outflow concentration (cfu or pfu mL −1 for microbes and mg L −1 for Br), D is the dispersion coefficient (cm 2 min −1 ), V is the average pore-water velocity (cm min −1 ), x is the transport distance (cm), and λ is a first-order temporal removal rate (min −1 ). As microbial tracers travel through different pore networks from that of the solute tracer Br, we consider that the pore-water velocity and dispersion of microbial transport are different from those for Br and were thus, estimated independently from Br. The CXTFIT optimization package, version 2 (Toride et al., 1995), was used for deriving values of V, D, and λ.
The degree of velocity enhancement is expressed as velocity enhancement factor (VEF):

Soil structure and macroporosity
Image data derived from the X-ray CT scans was used to analyze the soil structure, heterogeneity, and macroporosity of the cores ( Table 4). The three core types are typical of those found in the Canterbury alluvial aquifer system which are described more fully by Dann et al. (2009). They describe the alluvial gravel media as bimodal very poorly sorted, very finely skewed, leptokurtic coarse to very coarse gravels, and found that the sands are unimodal moderately well sorted, symmetrical, or finely skewed, platykurtic medium sands, with the highly permeable OFG lenses being unimodal moderately to poorly sorted, finely skewed, leptokurtic or mesokurtic, medium to coarse gravels. Particle size distribution analysis undertaken on samples from the sites showed that the average volumetric percentage of gravel was relatively consistent (70-73%) (Dann et al., 2009). OFG material had an average mean particle size of 17.3 mm with a narrow range of particle sizes and was well sorted (D 90 /D 10 = 5.0). Particle size distribution data from the SLs indicated well-sorted sand with a narrow range (D 90 /D 10 = 3.3) and an average mean value of ∼0.3 mm (Dann et al., 2009). Core 1 shows mixed SG (0-89 mm) over a distinct SL (indicated by the vertical double headed arrow) (89-244 mm) and another SG layer (244-400 mm). Core 2 consists of a heterogeneous SG mix (0-400 mm). Core 3 shows SG (0-281 mm) over an OFG layer (indicated by the vertical double headed arrow) (281-384 mm) and another SG layer (384-400 mm) (Figure 3). The larger cracks through core 1, as can be seen in Figure 3 through the top SG layer and SL, were filled with cooled petroleum jelly and, therefore, do not appear to have an influence on the results (Tables 2 and 3).

Saturated flow tracer experiments
Good matches between the experimental tracer data and fitted BTCs were observed for the saturated experiments for all three tracers (Figure 4). Mass recoveries for Br were 100% for cores 2 and 3 and slightly less (82%) for core 1 (Table 2). There was little removal of both microbes in core 3, which contained an OFG lens, compared with core 2, which con-tained only SG. Core 1, which showed presence of biofilm growth, had slightly lower mass recoveries for Br (82%) and MS2 phage (64%), and very reduced mass recovery for E. coli of 4% (Table 2; separate vertical axis in Figure 4). The BTCs for core 1 showed similar rising limbs for all three tracers with decreases first observed for E. coli and then MS2 and Br, indicating similar transport in the larger pores where the most rapid transport occurs and much more filtration occurring in the smaller pores, especially for the larger E. coli bacteria. The increased removal (lower mass recoveries) for E. coli and MS2 phage for core 2 compared with core 3 indicates that the SG is a better texture compared with OFG to attenuate microbes under saturated conditions, as would be expected. The BTCs in Figure 4b for core 2 show a double peak for MS2, and some variability of the peak shape for both Br and E. coli, which could be consistent with dual transport pathways. Multiple transport pathways have also been noted in heterogeneous alluvial gravel groundwater systems (Pang et al., 1998).

Unsaturated high flow (discharge velocity ∼10 mm h −1 )
There were good matches between the observed and fitted Br BTCs for the unsaturated high flow experiments, with relatively poorer fits for the microbes (Figure 5-note the variable y axes for the microbes). The poor fits related to the peak concentrations-one observation point for MS2 in Figure 5b and two observation points for E. coli in Figure 5c. As all datapoints had equal weights during CXTFIT model simulations, the model fitted most datapoints but not those datapoints with extreme values. Both E. coli and MS2 showed significant removal (peak C/C 0 = 0.01 and 0.03, respectively) and greater variability is expected at these low recovery rates. Mass recoveries for Br were around 100%, with lower mass recoveries for MS2 phage (0.4-36%) and even lower mass recoveries for E. coli (0.15-1.5%; Table 3). Greater recovery was observed Vadose Zone Journal F I G U R E 3 X-ray computed tomography (CT) spinal and coronal images of intact cores (approximately 400 mm in length). Spinal images taken from middle of each core in SL, SG, and OFG for core 1 (Image #360), 2 (Image #350), and 3 (image #385), respectively for both MS2 phage and E. coli in core 3 compared with core 1 and core 2, which can be attributed to the presence of the OFG lens in core 3, with core 2 (SG) showing the highest removal of E. coli and MS2 phage (Table 3). There were similar trans-port velocities for all three tracers for core 3, but transport of MS2 phage was faster for core 2 compared with the other tracers and E. coli was slower than the other tracers for core 1 ( Figure 5).

Unsaturated low flow (discharge velocity ∼0.5 mm h −1 )
Under the low flow conditions, the shape of the Br peak was generally rectangular with a flat sloping plateau ( Figure 6), with mass recoveries still around 100% (Table 3). The peak C/C 0 values ranged between 0.83 and 0.91. The broad Br BTC at low flow unsaturated conditions indicates transport through a range of pore sizes within the cores.
The mass recoveries for both microbes were very low for the unsaturated low flow experiments. There was complete removal of E. coli observed in cores 1 and 2 (Table 3), with only a single E. coli observed for core 3 at the detection limit of 1 cfu mL −1 (C/C 0 = 6.4 × 10 −6 ). There were a small number of detections for MS2 phage for all three cores at very low values for C/C o (Table 3). These detections all occurred near the start of the rising limb of the Br BTC (data not shown) indicating that the transport had occurred in the larger active pores that had the highest transport at these low saturation conditions. The much lower observed concentrations indicates that most transport was occurring in the very small pores as would be expected. Overall, the percentage mass recoveries for MS2 phage ranged from 7 × 10 −5 to 9 × 10 −3 F I G U R E 5 Breakthrough curves of the high flow (10 mm h −1 ) unsaturated cores with CXTFIT predicted curves for bromide (Br), Escherichia coli, and MS2 phage (Table 3) with core 2 showing the greatest removal (lowest mass recovery).

Modeling data
The transport parameters derived from the CXTFIT model for all flow conditions are shown in Table 5. The relative microbial mass recoveries show the substantial differences between the saturated and unsaturated flow conditions. If the saturated flow experiment for core 1 is omitted from the results (noted in next paragraph) then the relative mass recoveries for E. coli averaged 0.66, 0.002, and not detected for saturated, high flow unsaturated, and low flow unsaturated conditions, respectively. This can be summarized as minor removal for E. coli under saturated conditions, 2 orders of magnitude further removal under high flow unsaturated conditions and more than 3 orders of magnitude further removal between the high and low unsaturated conditions. Mass recoveries were slightly higher for MS2 phage and showed a similar pattern with relative mass recoveries averaging 0.86, 0.033, and 1.4 × 10 −5 for saturated, high flow unsaturated, and low flow unsaturated conditions, respectively (Tables 2  and 3). This equates to minor removal for MS2 phage under saturated conditions, 1 order of magnitude further removal under high flow unsaturated conditions and 3 orders of magnitude further removal between the high and low unsaturated conditions. The saturated flow experiment for core 1 showed much lower peak concentrations and mass recoveries (Tables 2  and 5). This was mainly observed for E. coli, where the mass recovery was between 10 and 20 times lower than for cores 2 and 3, but mass recoveries for MS2 phage and Br were also slightly lower than observed for cores 2 and 3 (Table 2). This was probably a result of biofilm growth (Section 2.6). The rising limb of the BTC for core 1 (Figure 4) was very similar for all three tracers but the peaks and tails were shifted to the left for the microbial tracers compared with Br, especially for E. coli, which would be consistent with a greater degree of clogging of the pores.
Core 2 had greater removal of both E. coli and MS2 phage compared with core 3 for the saturated experiments, and greater removal of MS2 phage in both high and low flow unsaturated experiments compared with both cores 1 and 3. The removal of E. coli was similar for all three cores in the high flow unsaturated experiments and there was complete (or nearly complete) removal of E. coli in all cores for the low flow unsaturated experiment (Table 3). For core 3, F I G U R E 6 Breakthrough curves of the low flow (0.5 mm h −1 ) unsaturated cores with CXTFIT predicted curves for bromide (Br). There was almost complete removal of the microbes under the low flow conditions (see Tables 3 and 5) there was one detection of E. coli at the detection limit for the low flow unsaturated experiment (Table 3).
The CXTFIT model fits for the Br BTCs were excellent (R 2 > 0.97) and more variable but mostly very good (R 2 > 0.90) for the microbial tracers (Table 5). Core 2, comprised entirely of SG media, had higher removal rates compared with cores 1 and 3 for the high flow unsaturated experiments. The relative velocities of E. coli compared with Br were <1 for both the saturated flow and the high flow unsaturated experiments whereas that relative velocities for MS2 phage relative to Br were more variable but closer to one for both the saturated and high flow unsaturated flow experiments (Table 5). This can also be clearly seen in Figures 4 and 5.

DISCUSSION
The vadose zone, being the area beneath the plant root zone but above the capillary fringe of the groundwater table, expe-riences variable saturation conditions. Near saturated flows in the vadose zone are likely to occur when field capacity of the overlying soil has been exceeded. Irrigation regimes such as flood irrigation are likely to meet these conditions as are spray irrigators when high volumes of water are applied or when they are applied directly before or after a natural rainfall event. Saturated flow conditions in the vadose zone also occur beneath onsite wastewater systems during times of high usage. In New Zealand, common spray irrigation technologies include center pivot and traveling irrigators. Center pivot irrigators have a typical application depth of 18 mm every 3-4 days, with instantaneous application rates ranging between 5 mm h −1 near the center and 80 mm h −1 at the end of the center pivot. The application depth for a traveling irrigator is typically 50-60 mm, with an instantaneous application rate of 10-40 mm h −1 (Close et al., 2010). Infiltration rates for soils typically range between 1 and 30 mm h −1 depending on the soil texture and these application rates for application and infiltration provide a context for the unsaturated flow rates of approximately 0.5 and 10 mm h −1 used in this study. While steady saturated flow in nature is rare it does represent a worst-case scenario for modeling purposes. Flows in more unsaturated conditions, while more complex to model and assess, are more representative of soil/vadose zone processes and looking at a range of unsaturated to saturated conditions provides important insights into microbial transport. There were significant changes in hydraulic conductivity observed with saturation status, with K averaging 4.3, 0.18, and 0.013 m d −1 , for the saturated, high, and low flow unsaturated flow rates, respectively. These changes were expected and can be predicted using relationships developed by Richards, Van Genuchten, and others as described in Van Genuchten & Simunek (2004). However, it was not possible to measure the saturation status accurately enough for these intact cores of SG to check the quantitative agreement with these relationships.
With a decrease in saturation status, the transporting pore network changes from all pores being filled and transporting water and contaminants under saturated conditions, to a smaller fraction of increasingly small pores being active in transport processes at lower saturation conditions (Engstrom et al., 2015). The decrease in the sizes of pore and throat diameters for the transporting pores has a compounding impact on the removal of the particulate microbes as compared with the dissolved Br tracer. This is clearly observed in these series of experiments, with relatively minor removal of microbes under saturated conditions, reductions of 2-3 orders of magnitude under the high flow unsaturated conditions, and complete or almost complete removal under the low flow unsaturated conditions (Table 3). The removal rates and velocities fitted from the CXTFIT model also show these differences in microbial transport and removal for the different flow conditions (Table 5).

Vadose Zone Journal
The BTCs for the saturated flow experiment in core 1, which showed similar rising limbs for all three tracers (Figure 4a), is an illustration of the size exclusion effect that is well known in heterogeneous saturated groundwater systems and has been described by Pang et al. (1998Pang et al. ( , 2005. This is where solutes such as Br will travel through the whole pore network whereas particles such as microbes will be restricted to the larger pores. Water is transported at higher velocities through the larger pores so the particles, traveling predominately through the larger pores, will be predominately observed on the rising limb of the BTC and travel, on average, at a higher velocity compared with the Br, which travels through the full range of the pore network. The faster transport of microbes than Br is reflected in the model fitted porewater velocity values and calculated VEFs (Table 5). It was noted that the C/C o was greater for MS2 (0.59) compared with Br (0.46) for the saturated flow experiment in core 3 (Table 2). BTCs for microbes are often narrower (lower dispersion) compared with solutes because the microbes are only transported through a fraction of the pore network as noted above. Pang et al. (1998Pang et al. ( , 2005 provide examples of narrower peaks and lower dispersion for microbial BTCs in alluvial gravel groundwater systems. As the Br BTC is more dispersed compared with the BTC for MS2 the C/C 0 values are lower. This is also related to the experimental conditions as with a longer injection time the C/C 0 values for both Br and MS2 would be expected to reach a value of 1.0 given 100% mass recovery for both tracers (Table 2). Core 3 had the OFG lens within the core thus recovery of all tracers was greatest for this core under saturated conditions.
The inactivation rates determined from the injection mix were 0.026 d −1 for E. coli J6-2 and 0.246 d −1 MS2 phage. The saturated flow experiments (peak concentrations observed ∼1 h after injection) and high flow unsaturated experiments (peak concentrations observed 6-7 h after injection) were sufficiently short so that they were not significantly affected by these inactivation rates. The low flow unsaturated experiments had Br peak concentrations observed between 2 and 3 days. At 3 days, the measured inactivation rates would have resulted in a decrease of microbial counts 92% for E. coli J6-2 and 48% for MS2 phage. The maximum observed counts for the low flow unsaturated experiments were 6.4 × 10 −6 for E. coli J6-2 and 4.1 × 10 −4 for MS2 phage (Table 3), indicating that inactivation was a minor factor in the observed decreases.
The relative transport of bacterial compared with viral microorganisms can vary widely (Pang, 2009). Theories behind the differing transport rates have been suggested as being predominantly related to size, motility, and surface properties of the microorganisms (Engstrom et al., 2015;Schijven et al., 2017). In this study E. coli J6-2 is a rod-shaped bacterium with a diameter 40 times the size of MS2 phage (the viral model chosen) and a volume 106 times higher. In this study, E. coli was transported with some removal through all three intact cores under saturated conditions. The greater removal (lowest mass recovery) of E. coli compared with MS2 phage in the saturated flow experiments (Table 2) can be related to this difference in size. The lower removal of both the E. coli and MS2 phage in core 3 was attributed to the presence of the OFG which has much less fines and higher porosity.
Under high flow unsaturated conditions (i.e., 10 mm h −1 ), the amount of E. coli recovered at the bottom of the cores was significantly reduced and E. coli showed greater removal from the cores compared with MS2 under both low and high flow unsaturated conditions. There was greater removal of MS2 phage for core 2 (SG) compared with the cores with OFG and SL in the high flow unsaturated experiments (Table 3). This suggests that more microbial removal occurs in the SG material. Previous investigations have shown that the SG is tightly packed with much lower mean porosity (17%) compared with the porosities observed for the SL and OFG (32-34%) (Dann et al., 2009). Under the unsaturated low flow regime (0.5 mm h −1 ), there was no E. coli measured in the leachate in cores 1 and 2 and only a single detection in core 3 (Table 3). The presence of the OFG or SL did not appear to affect the transport of E. coli through the cores. However, given the very low recoveries seen for the low flow unsaturated experiments, we cannot be sure that many of the microbes had even reached the SL or the OFG.
Flow characteristics and water content appear to be the biggest predictors of microbial transportation risk. Studies have shown the persistence of microorganisms in drier soils (Cools et al., 2001;Mubiru et al., 2000). Previous studies have also shown that virus transport in a sand-filled lysimeter is more attenuated under unsaturated compared with saturated flow conditions (Chu et al., 2003;Powelson & Gerba, 1994;Powelson et al., 1990). In this study, microbial removal was much greater under unsaturated flow conditions than under saturated flow conditions. This is indicated from the peak concentrations, mass recoveries, spatial removal rates (Tables 2  and 3) and model-derived temporal removal rates (Table 5). Suggested mechanisms for microbe removal under unsaturated conditions are straining or filtration by the media or attachment to the air-water interface or thin films of water (Flury and Aramrak, 2017;Jiang et al., 2008).
The results of this study appear to be consistent with those of Weaver et al. (2016), Jiang et al. (2008), Powelson & Mills (2001) and Balkhair (2017). Weaver et al. (2016) showed that when flood irrigation (i.e., saturated flow conditions) was applied to variously aged cowpats, E. coli was readily transported down through a gravel core and had the potential to contaminate groundwater. In the study by Jiang et al. (2008), bacterial leaching increased with increasing water potential and water content, i.e., the wetter the soils, the greater the bacterial leaching. This is expected, as with an increase in wetness, the air-water interface effect reduced; and would result in greater bacterial leaching. Balkhair (2017) described sev-eral soil columns (i.e., not vadose zone which probably had significantly more organic matter than our gravel cores) and showed that flow under unsaturated conditions removed more E. coli than flow under saturated conditions. The irrigation rates used in Balkhair (2017) were 0.43 and 0.95 mm min −1 (equal to 26 and 57 mm h −1 , respectively), which are considerably higher than the rates of 0.5 and 10 mm h −1 used in this study. Under these conditions, the bacterial retention was about 42% and 50% for each soil and the retention was suggested to be a factor of soil properties but also the degree of saturation of each of the columns. Balkhair (2017) found that a reduction in water content reduces the peak concentration and, therefore, reduces the total recovery. The air-water interface was described as a possible site under unsaturated conditions where bacteria and possibly viruses are either retained or inactivated.
The flows through the intact cores under the unsaturated conditions are representative of drainage occurring in response to normal rainfall intensities. The flow rates are much slower, the pore spaces are variably filled with air and water and there is greater chance for the microorganisms to be retained. As observed in this current study, lower peak concentrations under unsaturated flow were also observed by Powelson and Mills (2001), who evaluated the transport of E. coli through sand columns under saturated, constant unsaturated and cyclic unsaturated flow.
Bacteria applied to columns under different suction regimes are known to enter pore sizes of different sizes. Sirivithayapakorn & Keller (2003) presented experimental evidence of the effect of colloid exclusion from areas of small aperture sizes and found that the smallest pore size to colloid ratio entered by a range of colloids (0.05-3 μm) was 1.5. It is expected that the sandy fraction of the SG and SL are where the microbial attenuation takes place. Dann et al. (2011) carried out micro X-ray CT analysis on intact drill cores containing medium sand fractions and measured distributions of pore and throat sizes. The observed ratios of mean throat to mean grain diameter were about 0.2, with the throat diameters ranging from 0.05 to 0.2 mm for the medium sand fractions. OFG would have much larger throat diameters (orders of magnitude greater) as there are very few fines. However, the presence of the OFG lens did not appear to affect the vertical transport of the contaminants, and the transport of the tracer and microorganisms through the vadose zone appeared to be controlled predominantly by the SG material. This is likely an experimental artifact as flow in the intact cores was constrained to the vertical direction by the dimensions of the column (Figure 2). In a real-world 3D situation, however, the OFG are more likely to contribute to significant lateral rather than vertical transport. Burbery et al. (2017) in a study examining the connectiveness of OFG, showed significant lateral movement of water under saturated conditions. Water was applied about 15 m from a sea cliff, and moved 28 m in the transverse direction, while moving vertically a distance of around 5 m. This illustrates the complex nature of OFG connections in the real world and the limitations of laboratory scale experiments. Under large drainage events, microbes can easily pass through the alluvial gravels (especially via lateral flow through OFGs) while in slow flow regimes the microbes are mostly removed.
Compared with the saturated flow experiments, microbial removal was 1-2 orders of magnitude greater in the high unsaturated flow experiments and a further 3-orders of magnitude removal between the high and low unsaturated experiments (Tables 2 and 3). The much greater removal of microbes under unsaturated conditions have significant implications for land management decisions. The high flow unsaturated conditions tested in this study of 10 mm h −1 are high infiltration rates yet result in 1-2 orders of magnitude greater removal of microbes compared with saturated conditions. The low flow unsaturated conditions tested in this study of 0.5 mm h −1 are still reasonably high infiltration rates yet result in 4-5 orders of magnitude greater removal of microbes compared with saturated conditions. This clearly illustrates the substantial benefits in terms of microbial groundwater contamination of reducing infiltration rates through the vadose zone to below saturated conditions.

CONCLUSION
Retention of fecal microbes in vadose zone is highly dependent on flow rates and degree of saturation. Under saturated conditions, there is significant potential for microbes to be transported. The removal of microbes increases by orders of magnitude as flow rates and degree of saturation decreases. Different flow regimes resulted in different transport characteristics through the alluvial gravels; units with sand within them increased retention of microbes and increased dispersion of solutes. It should be noted that in a real-world 3D situation, however, the OFG are more likely to contribute to significant lateral rather than vertical transport and may be connected with other OFG lenses at scales greater than observed at the laboratory scale. From a land management perspective, the risk associated with the presence of saturated conditions in the vadose zone should be considered and avoided if possible, as retention and removal of microbes is orders of magnitude lower under saturated conditions.

A C K N O W L E D G M E N T S
This study was funded by New Zealand Ministry of Business Innovation and Employment under contracts C03×1001 and C03×1701. We would also like to thank Pieter Havelaar, Ann-Kathrin Eisenkolb, Julia Nowatschek, Bastian Knorr, and Hazel Clemens for their contributions in conducting some of the experiments.