Real-time characterization of flow and transport processes in the vadose zone, as well as assessment of chemical transformation of the percolating water during the remediation experiments, was carried out with a VMS that was installed across the entire unsaturated profile, from land surface to a depth of 37 m (Fig. 2). A detailed description of the VMS, its structure, installation procedure and performance can be found in previous publications (Dahan et al., 2009; Rimon et al., 2011a) and in the Supplement. In particular, the VMS that was used at this site was composed of a 44 m long flexible polyurethane sleeve hosting 11 monitoring units distributed along its length. Each monitoring unit included: (a) a flexible time-domain reflectometer (FTDR) sensor for continuous measurement of variations in the sediment water content (Dahan et al., 2008; Rimon et al., 2007), and (b) vadose zone sampling ports (VSPs), which enable frequent sampling of the vadose zone pore water for chemical analysis (Baram et al., 2012a; Dahan et al., 2009; Rimon et al., 2011b; Turkeltaub et al., 2016). The VMS flexible sleeve was installed in a 0.16 m diameter uncased borehole drilled slanted at a 55∘ angle (to the horizon) to a vertical depth of 37 m. In addition to the 11 monitoring units that were installed with the VMS, four additional monitoring units were installed directly in the soil at depths of 0.5 and 1.5 m. It should be noted that the slanted installation is preferred to ensure that measurements carried out by each monitoring unit take place in separate undisturbed sediment columns. In addition, the flexibility of the monitoring sleeve and its filling with non-shrinking cement grout ensured complete sealing of the borehole void and prevention of cross-contamination through preferential flow in the borehole.

Water amended with ethanol as the electron donor for perchlorate-reducing bacteria was infiltrated into the vadose zone through an area of 8 × 30 m at the pilot site using a drip-irrigation system. Dripping lines with drippers having a nominal discharge rate of 2.2 L h-1 were set up in a 0.3 × 0.3 m spatial distribution to create fairly even water distribution over the area. Accordingly, the total discharge rate of the irrigation system was set to 5 m3 h-1, which is equivalent to an infiltration rate of 0.02 m h-1. To promote anaerobic conditions in the unsaturated zone, a polyethylene liner covered with soil was placed over the dripper system after its installation. Ethanol was selected as the electron donor and carbon substrate because it is a natural, soluble compound that is commonly used by perchlorate-reducing bacteria (Bardiya and Bae, 2011). Moreover, it reduces potential increase in soil salinity associated with other common sources of electron donors such as acetate (Gal et al., 2008).

Three infiltration experiments with variable amounts of water and ethanol were implemented at the pilot site over a period of 7 months. To trace the percolating water across the unsaturated zone, bromide (as KBr) was added to the infiltrating water at the early stages of the experiment. The infiltration rates, as well as the concentrations and application sequence were assigned for each experiment with insight gained from the previous experiment (Table 2). Accordingly, information obtained by the VMS on depth propagation of the ethanol and tracer and variations in perchlorate and chloride concentrations across the unsaturated zone during and after each infiltration experiment were used to adjust the infiltration procedure in the following stage.

The first experiment (8 August 2010) consisted of infiltration of 50 m3 water (equivalent to 210 mm) (Table 2). The first 6 m3 were applied as untraced fresh water with no ethanol to wet the topsoil. This wetting stage is essential to promoting deep transport and preventing accumulation of tracers and ethanol in the low-flow zone located on the margins of the dripper's influential zone. Following the initial wetting phase, 0.4 m3 of bromide tracer solution (as KBr) at a concentration of 12.5 g L-1 was applied, followed by 1 m3 of water with 5 % ethanol. Immediately after the application of the carbon and tracer solution, the rest of the water (42.6 m3) was applied to enhance transport of the ethanol and tracers to deeper parts of the vadose zone.

After obtaining the results pertaining to the wetting process, as well as tracer and ethanol migration in the vadose zone during the first infiltration experiment, a second experiment was performed (1 September 2010). This experiment was conducted with 100 m3 of water (equivalent to 420 mm). Here the first 7 m3 of water was injected into the topsoil as untraced fresh water, followed by 1 m3 of water with 5 % ethanol, and then the rest of the water dose (92 m3). No tracers were used in this experiment. The amount of water used after application of the ethanol was doubled to enhance migration of the ethanol to deep sections of the unsaturated zone.

Results from the first two experiments indicated limited migration of tracer and ethanol to deeper parts of the vadose zone. A third infiltration experiment was therefore conducted 5 months later with increased discharge of 300 m3 (equivalent to 1250 mm). This experiment started with 24 m3 of untraced water followed by 0.4 m3 concentrated (50 %) ethanol solution. Then, the rest of the water (275.6 m3) was used to push the ethanol down into the vadose zone. The large quantity of water applied after the concentrated ethanol solution was designed to enhance quick migration of the ethanol to deep parts of the vadose zone while minimizing its biodegradation in the upper soil layers.

Perchlorate was analysed with a perchlorate ion-selective electrode (ISE; Laboratory Perchlorate Ion Electrode, Cole-Parmer, USA). All samples measured with the ISE were adjusted by dilution to a concentration range of 10–100 mg L-1. Duplicates were frequently analysed by injecting 25 µL sample into a Thermo Scientific^™ Dionex^™ ion chromatography system (ICS 5000) equipped with Ion Pac AS19 column (detection limit of ±0.01 mg L-1). Because results from the two methods were not significantly different, most of the data reported here are from the perchlorate electrode with a detection limit of 1 ppm. Bromide and chloride were analysed by ion chromatography with a detection limit of 30 ppb (Gal et al., 2008). Total organic carbon (TOC) was analysed to examine the success of delivering carbon to the vadose zone. Because porewater samples from the vadose zone are obtained through the VSP, which uses a porous ceramic interface (pore size < 2 µm), TOC values reflect DOC. TOC was analysed through a combustion TOC analyser (Teledyne Tekmar, Apollo 9000) with a detection limit of 2 ppm. Ethanol concentration in the vadose zone pore water was analysed in a gas chromatograph (Varian, CP3800). Water samples (1.5 µL) were injected by autosampler. The flame ionization detector and injector temperatures were set to 270 and 250 ∘C, respectively. The GC oven temperature was first held at 50 ∘C for 1 min, increased to 220 ∘C at a rate of 25 ∘C min-1, and then held for 4 min. The separation was performed by Stabilwax^® capillary column (60 m, 0.32 mm, 0.25 µm, Restek Corporation, USA); helium was used as the carrier gas (1 mL min-1). For quantification, five external standards were used.

Temporal variations in the vadose zone water content provide a direct indication of percolation processes in the vadose zone (Rimon et al., 2007; Dahan et al., 2008; Turkeltaub et al., 2015a). Each infiltration experiment launched a wetting wave that propagated sequentially through the unsaturated zone (Fig. 3). Down-migration of the wetting wave was expressed as a quick rise in water content followed by a recession caused by water redistribution and drainage. Referring the wetting sequence in the vadose zone to the infiltration events on land surface enabled a direct calculation of the flow velocity across the unsaturated zone (Rimon et al., 2007; Dahan et al., 2008). All three infiltration experiments produced wetting fronts that moved down the vadose zone at a velocity of ∼ 0.18 m h-1, even though the water volumes that were used in each experiment were significantly different (50, 100 and 300 m3). Additional information on calculation procedure of flow velocities may be found in the Supplement. Observations of regulated flow velocities at constant rates across the vadose zone under variable surface hydraulic conditions have also been reported in other studies (Dahan et al., 2008; Amiaz et al., 2011; Rimon et al., 2011a).

The high salinity of the deeper parts of the vadose zone (> 13 m) (Fig. 1) limits the reliability of the TDR technology for measuring water content at those depths (Nadler et al., 1999). Therefore, variation in water content, as an indication of deep percolation, is presented here only down to a depth of 11.2 m, where the salinity was low enough to achieve reliable moisture measurements with the FTDR sensors. Nevertheless, indications of deep percolation in the deeper layers (> 13 m) are further discussed through the variation in chemical composition of the percolating water across the entire thickness of the unsaturated zone (40 m).

Initial analysis of pore-water samples from the vadose zone, prior to initiation of the infiltration experiments, revealed very high concentrations of perchlorate and chloride, both reaching maximum values of ∼ 22 500 mg L-1 (Fig. 1), and total dissolved solids (TDS) of 43 000 mg L-1, at a depth of 21 m. Note that at this stage, the concentrations of perchlorate and chloride are nearly identical throughout the entire profile. These high concentrations, sampled by the VMS, are in accordance with concentration profiles obtained previously in extracts of sediment samples (Gal et al., 2009).

Frequent sampling of the vadose zone pore water showed dynamic variations in perchlorate concentration during the percolation experiments. In the upper section of the vadose zone (0–13 m), perchlorate concentrations decreased dramatically, from as high as 9000 mg L-1 to below detection levels (Fig. 4). Such a reduction in concentration in a relatively thick portion of the vadose zone (13 m) over the short period of 10 months is clearly desirable and may even be considered a great success. Nevertheless, closer inspection of the variations in perchlorate concentration in deep parts of the vadose zone (17–40 m) showed an increase at most of the measurement points (Fig. 5). Perchlorate concentration rose from 12 700 to 27 400 mg L-1 at a depth of 17 m during the same period. A similar increase in concentration was also found in deeper parts of the cross section at depths of 25, 28, and 36 m. Note that during this period, an increase in perchlorate concentration was even observed in the groundwater (represented at a depth of 41 m in Fig. 5). Obviously, the mixed trend in variations of perchlorate concentration implies that transformation and mobilization processes take place simultaneously. As such, the conditions for both biodegradation and mobilization should be examined along with the variation in perchlorate concentration.

Available organic carbon as an electron donor is crucial for perchlorate reduction. To increase the concentration of DOC in the vadose zone, ethanol was mixed with the percolating water during the early stage of each infiltration experiment. Analysis of ethanol and DOC in the water samples from the vadose zone throughout the experiment revealed high correlation between the two. Theoretically 1 g per litre of ethanol is equal 0.52 g L-1 of soluble carbon. However, in the site the dissolved carbon is composed of ethanol, its microbial metabolism products (such as acetate), and other soluble microbial metabolites that can also serve as electron donors. Thus, DOC provides a better knowledge on the availability of electron donors in the soil pore water. Accordingly, we assume that the variation in DOC during the experiments was due to transport of ethanol or ethanol-degradation products with the percolating water (for further details correlation between DOC and ethanol concentration see the Supplement).

During the first infiltration experiment, an increase in DOC above background levels was observed only at shallow depths, down to 1.5 m (Fig. 6). No signs of increasing DOC were observed in the deeper parts of the cross section at this stage. Twenty-three days later, before initiation of the second infiltration experiment, DOC values had dropped back down to background levels. This implies that the ethanol was microbiologically consumed and mineralized to inorganic carbon in the soil before it could be leached further down.

As a result of the limited transport of electron donor, ethanol, in the first infiltration experiment, a second experiment was conducted with the same mass and concentration of ethanol. However, it was flushed with double the amount of water to promote its quicker migration to deeper layers (Fig. 4). In this experiment, no signs of increasing DOC were observed at any depth. However, DOC level decreased to values below background levels (Fig. 6). Obviously, the rate of ethanol metabolism and mineralization in the soil increased following the first experiment, where both water content of the sediment and substrate required for efficient microbial activity increased. As a result, ethanol-degradation efficiency in the topsoil (< 0.5 m) was significantly enhanced.

To overcome the limitation of electron donor delivery through the shallow soils, a third infiltration experiment was designed. In this experiment, the ethanol was injected in a 0.4 m3 high-concentration (50 % vol) pulse followed by a large volume of water. Application of ethanol at a very high concentration was aimed at suppressing its biological degradation in the shallow soil. The ethanol pulse was introduced after application of 24 m3, the latter to provide high initial wetting conditions under the ethanol front. Then the ethanol slug was pushed down with 276 m3 of water. At this stage of the study, which was conducted 6 months after the previous one, a substantial increase in DOC was observed in the entire top 13 m of the cross section (Fig. 6). Obviously, an increase in DOC serving as electron donor is an essential prerequisite for perchlorate degradation. Apparently, application of ethanol at a high concentration, which inhibited its degradation in the upper layer, succeeded to drive the ethanol all the way down to 13 m, just above the clay layer. Nevertheless, no signs of DOC increase were observed below 13 m.

The mechanism controlling down-propagation of a non-conservative substance such as ethanol may be elucidated by looking at the migration pattern of a conservative tracer such as bromide. Bromide was injected with the percolating water in the early stages of the first infiltration experiment. Results on bromide migration are presented here only for the top 13 m, where the background concentrations prior to the initiation of the infiltration experiment were below detection limits. Concentration profiles during the infiltration experiments clearly demonstrated sequential progress of the percolating water across the top 13 m of the unsaturated zone (Fig. 7). Mass balance calculation of bromide on the basis of the concentration profiles (Fig. 7) and sediment water content (Fig. 3) on various dates after the infiltration experiment resulted in high recovery rates of 85–127 %. A comparison of the transport patterns of bromide and DOC confirmed that biodegradable material such as ethanol is rapidly consumed in the vadose zone.

An increase in chloride concentration in the vadose zone is usually attributed to evaporation processes near land surface, a mechanism that is unlikely to occur in this particular setup where the surface is isolated from the atmosphere. Accordingly, variations in chloride concentration across the vadose zone may be attributed to chloride mobilization with the percolating water and perchlorate reduction. Therefore, degradation of perchlorate is expected to result in an increase in chloride mass.

Prior to the infiltration experiments, chloride-to-perchlorate ratios in the vadose zone were very similar, exhibiting nearly identical profiles (Fig. 1) with equivalent concentration proportions of 2.4–5.5 (Fig. 8). Following the infiltration experiment, a significant increase in ionic ratios was observed in the top 13 m, while in the rest of the profile – from a depth of 17 m to the water table – the concentration ratio of chloride to perchlorate remained relatively unchanged. Obviously, since both perchlorate and chloride are very soluble and mobile, infiltration water with a low concentration of chloride (∼ 100 mg L-1) and zero perchlorate is also expected to result in an increased chloride-to-perchlorate ratio, even if no perchlorate degradation takes place. Since both chloride and perchlorate are very mobile and easily displaced with the percolating water, quantification of the perchlorate-degradation rate with respect to its down-leaching is not straightforward.