Combining Geoelectrical Measurements and CO 2 Analyses to Monitor the Enhanced Bioremediation of Hydrocarbon-Contaminated Soils: A Field Implementation

Hydrocarbon-contaminated aquifers can be successfully remediated through enhanced biodegradation. However, in situ monitoring of the treatment by piezometers is expensive and invasive and might be insufficient as the information provided is restricted to vertical profiles at discrete locations. An alternative method was tested in order to improve the robustness of the monitoring. Geophysical methods, electrical resistivity (ER) and induced polarization (IP), were combined with gas analyses, CO 2 concentration, and its carbon isotopic ratio, to develop a less invasive methodology for monitoring enhanced biodegradation of hydrocarbons. The field implementation of this monitoring methodology, which lasted from February 2014 until June 2015, was carried out at a BTEX-polluted site under aerobic biotreatment. Geophysical monitoring shows a more conductive and chargeable area which corresponds to the contaminated zone. In this area, high CO 2 emissions have been measured with an isotopic signature demonstrating that the main source of CO 2 on this site is the biodegradation of hydrocarbon fuels. Besides, the evolution of geochemical and geophysical data over a year seems to show the seasonal variation of bacterial activity. Combining geophysics with gas analyses is thus promising to provide a new methodology for in situ monitoring.


Introduction
Petroleum hydrocarbon leaks and accidental spills happen commonly during the production, refining, transport, and storage of petroleum. Release of petroleum hydrocarbons into the environment causes damage to ecosystems [1] and to soil and water resources [2]. Increasing demand for drinking water and cropland with population growth requires effective remediation techniques to treat the contamination and to decrease hostile effects on health and environment. In situ remediation techniques such as enhanced bioremediation were shown to be effective in cleanup of the contamination [3,4]. Enhanced bioremediation involves the addition of nutrients or electron acceptors to the subsurface environment to accelerate the natural biodegradation processes which degrade hydrocarbons [5]. Due to the potential of cost saving of in situ techniques compared to conventional ex situ techniques, there is an economical interest for commercial providers to use enhanced bioremediation [6]. However, these processes remain partially unexploited, mainly because their in situ monitoring, before, during, and after soil treatment operations, is often expensive and technically challenging. Indeed, where significant subsurface heterogeneity exists, conventional intrusive groundwater sampling campaigns can be insufficient to obtain relevant information as they are restricted to costly monitoring piezometers at discrete locations. New monitoring tools are needed to overcome these limitations and make the enhanced bioremediation more reliable and robust, as well as economically competitive.
Previous studies suggest that geoelectrical techniques, especially electrical resistivity (ER) and induced polarization (IP), can be used to detect the presence of LNAPLs (Light Nonaqueous Phase Liquids) [7][8][9][10] as well as monitoring the effects of their biodegradation [11][12][13][14]. Biodegradation processes modify ground electrical properties because they change biophysicochemical conditions in the subsurface: (i) bacteria modify local redox conditions, inducing changes in self-potential (SP) [15]; (ii) microbial activity can produce organic acids and/or carbonic acid that affect the pore water conductivity, modifying both the in-phase and the quadrature conductivity [16]; (iii) during microbial growth and formation of biofilms, biomass can clog pores and potentially change the porosity and hydraulic conductivity, increasing the storage of electrical charges [17,18]. Thus, ER and IP are expected to be effective nonintrusive tools to monitor enhanced bioremediation.
Some studies have implemented geoelectrical methods to characterize hydrocarbon-contaminated sites under biodegradation [19][20][21][22], but only few apply these methods for long-term monitoring or to prove the efficiency of enhanced bioremediation [23]. These previous field studies suggest that ER and IP are highly sensitive to the biophysicochemical processes associated with biodegradation. Nevertheless, the interpretation of geophysical data remains challenging partly because several factors may contribute to the observed electrical response (presence of metallic particles or clays, e.g.) and/or influence the electrical response in function of time (variation of water saturation and temperature). That is why geoelectrical methods are often used in conjunction with geochemical measurements (temperature, pH, redox potential, water conductivity, and dissolved oxygen content) to detect changes in the chemical and physical properties of the soil and groundwater. With the aim of minimizing the use of piezometers, microbial activity is followed by studying gas emissions at the ground surface. Indeed, aerobic degradation of hydrocarbons results in the production of CO 2 [24]. Due to its limited solubility (in alkalinity saturated groundwaters), CO 2 gas will tend to migrate toward the ground surface and surface CO 2 fluxes can be directly linked to the biodegradation intensity. Meanwhile, this CO 2 production occurs concurrently with natural root and microbial soil respiration. Thus, tools capable of quantifying CO 2 sources are also needed. Carbon stable isotopic analyses (ratio of 13 C/ 12 C) use the fact that petroleum hydrocarbon isotopic signature is distinguishable (range from −18 to −34‰ versus Vienna Pee Dee Belemnite or VPDB [25]) from other aquifer components: for example, carbonate minerals have a signature between +2 and −12‰ versus VPDB [26] and the C 3 plants (plants that use the Calvin or C 3 cycle of carbon fixation) produce organic matter with isotopic signature near −25 ± 5‰ versus VPDB [25,27,28]. Moreover, bacteria can induce a carbon isotopic fractionation as they preferentially use the lighter isotope 12 C (due to the lower energy required to break intramolecular bonds) [29]. Some studies linked the CO 2 production with contaminant degradation rate in laboratory [30][31][32] and at contaminated field sites [33]. Some researchers have examined the isotopic signature of CO 2 and dissolved inorganic carbon from aerobic biodegradation of petroleum hydrocarbons to support their analysis [34][35][36][37][38]. Results reflect the difficulty in definitively attributing CO 2 produced from biodegradation of contaminants versus that from possible interferences of background CO 2 under field conditions. The other natural sources of CO 2 , plant root respiration in shallow sediments and carbonate dissolution in subsurface sediments, must be taken into account in field experiments [25].
Integrating both geoelectrical methods and gas analyses could help to develop a less invasive technique with a little geochemical sampling in piezometers in order to monitor enhanced biodegradation of hydrocarbons. These tools were previously tested at laboratory scale in sand-columns with toluene as model pollutant and an exogenous bacterial strain [39]. The promising results from the column study allowed implementing a monitoring strategy at the field scale. The site was a gasoline station where gasoline and diesel fuels leaked eighteen years ago. A reactive barrier supplied the necessary oxygen to the aquifer in order to stimulate the aerobic bacterial processes since April 2014. Geoelectrical measurements and ER and IP surveys, as well as CO 2 analyses, had been performed from February 2014 until June 2015 on this site.

Materials and Methods
The experimentation had been implemented at a gasoline station over a 16-month period, from February 2014 to June 2015. After a preliminary characterization of the site (baseline, February 2014), and the start of the oxidative reactive barrier in April 2014, the long-term monitoring had been carried out.
The site had been equipped to monitor time-lapse changes in electrical properties of the subsurface with electrical resistivity (ER) and time domain induced polarisation (TDIP) systematic 2D tomographies. This monitoring had been combined with regular CO 2 fluxes measurements and its carbon isotopic ratio at the ground surface.
The indirect monitoring was completed with several analyses in piezometers to assess the nonintrusive survey: (i) measurements of pollutant concentration and carbon isotopic signature of selected aromatic hydrocarbons and (ii) monitoring of groundwater physicochemical properties with probes.

Site Description.
The site studied was a gasoline station where gasoline and diesel fuels leaked in 1997 ( Figure 1). It is still in activity but the former tank installation (source of contamination) was dismantled. Several piezometers had been drilled for previous studies on this site. These studies showed that (i) the ground was mainly composed of silts and clays: a first layer (thickness ranging from 0.5 to 3.5 m) of backfills made of brown silts, which often contained gravels, followed by a succession of loamy strata, with varying degrees of sands, and even sandy clays with limestone or millstone blocks (until 5 to 7.5 m depth); and then argillaceous limestones are interbedded with thin layers of sandy clays; (ii) the ground had a low permeability (between 10 −6 and 10 −7 m/s); (iii) the direction of flow was northwest; (vi) there were indications of a natural attenuation (hydrocarbon degradation) which prove the presence of a bacterial flora, such as absence of dissolved oxygen, that matched the BTEX plume in water [40].
An important limiting factor in bioremediation of soils contaminated with hydrocarbons is the lack of oxygen to support microbial activities [41,42]. Injection of aqueous H 2 O 2 was the selected method to overcome this limitation and to stimulate aerobic metabolic processes. A permeable reactive barrier of 4 m depth × 35 m long × 1.5 m wide was implemented to stop the plume migration ( Figure 2). The site was equipped with 3 pumping wells just upstream the barrier, 3 injection wells in the barrier, and 2 injection wells just downstream the barrier. Pumped water was filtered through active charcoal. After filtration, diluted H 2 O 2 was added to the treated water and oxygenated water was reinjected in the barrier, through the 3 injection wells, slotted from −2 to −4 m in order to deliver oxygen in

Electrical Geophysical Measurements.
Geophysical methods are a standard tool for obtaining information on volumetric distributions of subsurface physical properties of rocks and fluids. Among several electrical methods which measure electrical properties of the ground, two of them were used: the electrical resistivity (ER) and the time domain induced polarization (TDIP). In a typical ER measurement, four electrodes are used. ER instrumentation measures potential differences between pairs of electrodes, where the potentials result from a current applied by the instrumentation between two other electrodes. By making measurements with current and potential electrodes at many different locations, one can collect sufficient data which allow the construction of an electrical resistivity section (a 2D slice of the earth).
Whereas ER is sensitive to pore fluid resistivity (and consequently to total dissolved solids or ionic strength), the IP method measures the capacitive behaviour of the subsurface, which strongly depends on the properties of the mineral surface. Different studies have shown the sensitivity of IP to mineral precipitation and changes occurring at pore-grain interfaces, where biological reactions occur [12]. TDIP measurements use the same instrumentation as ER measurements. Here, the injected current is alternately ON and OFF. The decrease of the induced voltage is measured to calculate the chargeability, which is the ground capacity to store charges. Hereafter, in this paper, the normalized chargeability (defined as the chargeability divided by the resistivity magnitude) will be calculated to free the chargeability from variation of resistivity [43].
Two time domain ER and IP profiles were carried out at the study site. Positioning of the profiles was performed according to the contaminant plume and barrier location: both profiles went across the plume, one was perpendicular to the barrier, and the other was parallel, 8 m downstream the barrier ( Figure 2).
The perpendicular profile was buried permanently at 50 cm of depth. Apparent resistivity and chargeability measurements were obtained using a system with 60 stainless steel electrodes: 30 electrodes, with a constant spacing of 4 m, were dedicated to current injection, and the other 30, with a spacing of 4 m as well, were used for potential measurement to avoid electromagnetic (EM) coupling effect. The two sets of electrodes are disposed staggered, so that each electrode is laid out at a constant interval of 2 m. To achieve a good lateral resolution and to avoid EM coupling effect, the dipole-dipole array was applied to obtain the apparent resistivity and chargeability pseudosections. A Syscal R1+ (from IRIS Instruments©, Orleans, France) has been used in this study. Measurements were performed continuously every two days. The second profile, parallel to the barrier, was a temporary profile. It consists in 48 electrodes separated by 1 m. Measurements were performed every two months with an Elrec Pro (IRIS Instruments, Orleans, France).
After performing time-lapse ER and IP measurements, the electrical data were analysed in three stages: (i) Initial anomalous data were filtered by a process of elimination: (a) current < 100 mA; (b) potential < 0.1 mV; (c) apparent chargeability < 0; (d) error > 5% (resistivity measurements) and > 2% (chargeability measurements); (ii) The quadrupoles in common to all the data sets were selected; (iii) Data processing was performed with RES2DINV software [44], in time-lapse mode. The process is based on several iterations comparing the measured data with a model calculated by the software for each time series data set. This process enables obtaining a 2D distribution of resistivity and chargeability sections at every time. The following settings were used to define the type of constrains applied during the time-lapse inversion process: (a) The relative importance given to minimize the difference between models at different times is controlled by the time-lapse damping factor. A value of 0 to 5 can be generally used: if a value of 0 is used, the inversion of the different time series data sets would be carried out independently; if a value of 1 is used, equal weight would be given to reducing the difference between the models at different times; a larger value of the damping factor forces the different time models to be more similar. After some previous tests, a value of 3 was selected. (b) Time-difference roughness filter was set for "smooth changes" between time models. (c) The maximum number of iterations was 5.

CO 2 Analyses: Emissions Fluxes and Carbon Isotopic
Ratio. The geoelectrical measurements were combined with CO 2 analyses. The measurements of CO 2 effluxes at the ground surface demonstrate the occurrence and rate of aerobic hydrocarbon biodegradation [33]. However, to determine the accuracy of the method, contaminated-derived soil respiration must be distinguished from chemistry of mineralcarbonates and from the natural soil respiration, resulting from biodegradation of dead organic matter and plant root respiration.
CO 2 produced by fossil hydrocarbon degradation may be distinguished from that produced by other processes based on the carbon isotopic compositions, characteristic of the source material, and/or fractionation accompanying microbial metabolism. Indeed, carbon isotopic signature of CO 2 reaches those of contaminant and it can be accompanied by significant carbon isotope fractionation [37].
Stable isotopic ratio represents the abundance of the rare isotope with respect to the abundant isotope ( 13 C/ 12 C). Due to slight analytical variations, it is common to compare isotopic ratios measured in an unknown sample to those in a standard material; this results in the delta notation: with ( 13 C/ 12 C) VPDB = 0.011237 [45].
CO 2 concentration and 13 C isotopic ratio in produced CO 2 are measured by a high-resolution laser infrared spectrometer, called SPIRIT (Spectrometer Infrared In Situ Tropospheric) and developed by the LPC2E (CNRS, Orleans, France). This spectrometer is an adaptation of the original SPIRIT, described in   [46] and used for greenhouse gas emissions at the air-land interface [47,48], by the implementation of a quantum cascade laser using the 12 CO 2 and 13 CO 2 ro-vibrational lines. The uncertainty for 13 C determination on field deployment is on the order of 0.4‰. A complete description of the adapted SPIRIT for 13 C/ 12 C isotopic ratio quantification is available in Guimbaud et al. [49]. The spectrometer is directly connected with a closed accumulation chamber set on a permanent PVC cylinder collar (diameter = 30 cm) sunk into the soil to measure CO 2 emission fluxes and 13 (CO 2 ). The ground surface inside the PVC collar was cleared of vegetation (grass, roots) over 15 cm depth. We had 21 measuring points on the site, upstream, above, and downstream the reactive barrier ( Figure 2). Measurements were performed every two months since February 2014.

Geochemical Monitoring in Piezometers.
To assess the indirect monitoring of the enhanced bioremediation, chemical analyses were performed every two months in water samples from monitoring piezometers. BTEX concentration and carbon isotopic ratios of these BTEX were measured. Moreover, some piezometer probes had measured continuously the water table level, temperature, conductivity, and oxygen rate in monitoring piezometers (Table 1).

Baseline Characterization.
Baseline characterization consisted of collecting datasets prior to the oxygen injection (February-March 2014). Figure 3 shows the results for the permanent profile in March 2014. This profile went through the barrier (metric point 51 m) and was within close proximity to several piezometers (from north to south: Pz12, P5, Pz7, P3, Pz24, P1, and Pz4) whose positions are plotted on Figure 3. The temporary profile was located around 8 m downstream the barrier and it overlapped with two piezometers (from west to east: Pz24 and P2, Figure 4). The electrical resistivity tomography from the permanent profile showed resistivities between 40 and 80 Ω⋅m, over the 6 first meters, which correspond to the sandy-loamy layer of the aquifer. From 6-meter depth, resistivity was much smaller (<20 Ω⋅m) and this may be associated with a sandy clay level. The tomography from the temporary profile displayed the same resistivity range since the pilot site is quite homogeneous. Nevertheless, areas less resistant (<30 Ω⋅m) Regarding the normalized chargeability tomographies (chargeability divided by resistivity), the values were in the range of 0.025 to 1 mS/m, and the zones of higher normalized chargeability (>0.2 mS/m) matched the lower resistivity zones. These normalized chargeability anomalies correspond to a polarization effect. On the permanent profile, the normalized chargeability was quite high beyond 6 m depth. This confirms the presence of clays, very chargeable, at these depths. According to the geochemical data (Figure 1), the localization of these anomalies (higher normalized chargeability and lower resistivity) matched the hydrocarbon polluted area, downstream the former tank installation, at the top of the water table, where natural biodegradation produces conductive byproducts and where chargeable bacteria are numerous (proximity of an organic carbon source). Figure 5 displays the maps of measured CO 2 emissions and its carbon isotopic ratio in February 2014. There were quite high CO 2 emissions on this site, around 3.8 mol⋅m −2 ⋅s −1 , despite the fact that, during cold periods (December-March), gas fluxes are presumably very low [50,51]. These results were consistent with those found by Sihota et al. (2011) [33] where the average CO 2 efflux associated with contaminant degradation in the hydrocarbon-contaminated zone was estimated at 2.6 mol⋅m −2 ⋅s −1 . Moreover, the measured 13 C(CO 2 ), ranging between −22.2 ± 0.4‰ and −28.8 ± 0.4‰ versus VPDB (averaging −25.3 ± 3‰), were very negative compared to 13 C of atmospheric CO 2 (around −11‰ versus VPDB, one meter above the ground) and distinguishable from 13 C of CO 2 from vegetation respiration (around −28‰ versus VPDB [25,27,28]). This was an indication of natural biodegradation of hydrocarbons. Indeed, during biodegradation, CO 2 has an isotopic signature near those of the carbon source, which is quite negative in the case of fossil organic carbon (range from −23 to −29‰ versus VPDB for the BTEX of the site [40]). Figure 6 presents the BTEX (benzene, toluene, ethylbenzene, and xylenes) concentrations over time for 4 piezometers, between February 2014 (two months before barrier activation) and April 2015 (one year after barrier activation). There was an increase in the pollutant concentrations due to the remobilization of contaminants during the barrier construction. The discharge of BTEX may have had significantly slowed down the oxygen diffusion downstream the barrier: it had been quickly consumed and the amount of H 2 O 2 could not be increased for avoiding bacteria stress. And, one year after the beginning of biostimulation, there was just a little decrease in benzene and toluene concentrations.

One Year of Electrical Monitoring.
Before barrier activation, it was possible to localize a zone of natural biodegradation on the permanent and the temporary profile (Figures 3  and 4). Figure 7 shows the evolution of the electrical data over a year of enhanced biodegradation, for the permanent profile. The resistivity model sections were obtained after time-lapse inversion of 12 datasets, spaced about a month, acquired between March 2014 and April 2015. The convergence of iterative inversion is excellent (<1% for resistivity and <2 for normalized chargeability) and it is comparable with the noise level of the resistivity meter (1%, according to Iris Instruments). In order to compare the electrical data taking into account the impact of temperature and water saturation on electrical resistivity and chargeability, the inversed data were considered only in the saturated zone and adjusted at 25 ∘ C thanks to a linear law [52,53]: where is resistivity at temperature ( ∘ C), 25 is resistivity at 25 ∘ C, and ( ∘ C −1 ) is a temperature compensation factor. The temperature compensation factor was calculated based on Pz4 (located in a nonpolluted area) temperature data and found to be equal to 0.014, which means that resistivity decreases by 1.4% when temperature increases by 1 ∘ C. Figure 7 shows the anomaly of lower resistivity and higher normalized chargeability on all the tomographies, from March 2014 to April 2015. It was possible to identify two periods of decreasing resistivity and increasing normalized chargeability: a first period from March 2014 to October 2014, and a second period from March 2015 to April 2015. In order to quantify the evolution of the differences, the percentage difference of resistivity and normalized chargeability from March 2014 were calculated (Figure 8). A decrease in resistivity up to 15% is noticed in the area of interest (between metric points 50 and 70 m) at the end of the summer of 2014. However, such a seasonal variation cannot be seen on water conductivity (Figure 9(a)). Regarding normalized chargeability, the tendency is less clear and there are increases outside the area of interest as well.

One Year of CO 2 (Fluxes and Carbon Isotope Ratio)
Monitoring. Gas monitoring was carried out every onetwo months, between February 2014 and June 2015. The time allowed for measurements was very short (2 days per campaign). As a consequence, priority was given to flux measurements and then to isotopic ratio measurements of metering points with the highest emissions (more accurate). CO 2 fluxes ranged between 2.5 and 100 mol⋅m −2 ⋅s −1 (Figure 10(a)). As there was no vegetation inside the PVC changes in fluxes are supposed to be only due to temperature and moisture conditions. The difference of flux with season at this point was subtracted from the values obtained at the other measuring points to keep only the flux variations due to bacterial activity (degradation of natural organic matter and hydrocarbons) ( Table 2). A seasonal variation of the CO 2 emissions is still monitored (Figure 11): there were more emissions during summer and fall, when groundwater level is low (Figure 9(c)), water temperature is warmer (Figures 9(b) and 10(b)), and ground is less moist (Figure 10(c)). Moreover, CO 2 emissions had been measured with an isotopic signature typical of hydrocarbon biodegradation. Indeed, 13 C(CO 2 ) were measured between −27 and −30‰ versus VPDB since February 2014 (Figure 12(a)), in accordance with the BTEX 13 C signature measured in water (around −25 to −26‰ versus VPDB) (Figure 12(b)). Besides, CO 2 released at the surface originates from a more or less degraded BTEX located below the gas measuring station. Indeed, in December 2014, average evolution of 13 C(benzene) increased from (−28.1 ± 0.5)‰ to (−26.7 ± 1.2)‰ and 13 C(toluene) increased from (−27.4 ± 0.4)‰ to (−24.8 ± 1.9)‰, from upstream to downstream the pollution plume. At the same time, 13 C(CO 2 ) increased, from upstream to downstream as well, from (−30.5 ± 0.7)‰ to (−28.3 ± 1.6)‰ in July 2014 and from (−28.1 ± 1.1)‰ to (−27.0 ± 1.2)‰ in October 2014 [49]. Although these 13 C(CO 2 ) values were only a few per mil less than the 13 (CO 2 ) from microbial degradation of plant material, 13 (CO 2 ) showed a light isotopic fractionation (1 to 3‰) through the stream of the pollution plume: (i) 13 (CO 2 ) was lower than 13 C(hydrocarbon source).
(ii) The pollutant was enriched in 13 C when it was degraded, when residual BTEX concentrations were lower than 2000 g⋅L −1 .
(iii) The CO 2 tended to be depleted in 13 C, when fluxes were higher than 20 mol⋅m −2 ⋅s −1 .
It means that molecules with light isotopes ( 12 C) were preferred by bacterial metabolism. This is a typical isotopic fractionation due to biodegradation. This is in agreement with one example of a successful field application of isotopic assessment of biodegradation in an aviation gasoline contaminated aquifer where aerobic biodegradation of the gasoline constituents produced significantly enriched CO 2 [54]: the 13 C(CO 2 ) values were ranging from −22.0 to −26.3‰, soil organic matter for the site ranged from −13.5 to −26.1‰, and the aviation gasoline 13 C values were between    [54] concluded that the difference of 13 C was discernible. Another study from Aelion et al. (1997) [55] was carried out at a gasoline contaminated groundwater site under aerobic degradation. Although some of the 13 C(CO 2 ) values from the contaminated site (between −35.9 and −22.0‰) overlapped with background 13 C(CO 2 ) (between −22.9 and −23.9‰) from uncontaminated areas, on average the stable isotopic values were distinguishable between the contaminated and background sites.

Integration of Results and Conclusion
At the pilot site, the combination of geoelectrical measurements (ER and TDIP) with CO 2 analyses (fluxes and carbon isotopic ratios) was tested as monitoring tools to follow an enhanced biodegradation of hydrocarbons.
The results from the baseline characterisation in February-March 2014 had shown a more conductive and chargeable zone, around 3 meters in depth, which matches the polluted zone defined by geochemical borehole analyses, at the top of the water table (Figures 1, 3, 4, and 13). In this area, it can be assumed that there are numerous chargeable bacteria and production of conductive metabolites due to the pollutant degradation. This is consistent with the age of the gasoline leakage (almost twenty years ago), sufficient to allow the development of natural bacterial flora able to degrade hydrocarbons [56,57] and to transform an electrical resistive substance in a more conductive one [12]. Moreover, CO 2 emissions were strong in this area and had an isotopic signature typical of hydrocarbon biodegradation. However, the overlap between 13 C(CO 2 ) values from uncontaminated areas (plant root respiration and organic matter degradation) and from contaminated areas (with petroleum hydrocarbon degradation in addition) suggests that the combined measurement of 13 C(CO 2 ) and 13 C(BTEX) may be needed to confirm that the CO 2 is indeed from the contaminant degradation and not from microbial metabolism of natural plant materials. The northern zone of the site must be considered as a potential contaminated zone as well, accepting that the CO 2 emissions (measuring points 1, 12, and 16) are a sign of biodegradation and not the result of the presence of gas preferential pathways in this area ( Figure 13).
The treatment for enhanced biodegradation was started in April 2014, with the activation of the reactive barrier that brought oxygen to the water table to stimulate aerobic biometabolism. The regular monitoring over one year showed that the area of biodegradation was still detected by both indirect methods ( Figure 13). Moreover, a seasonal variation (after correction of effects associated with groundwater and soil temperature and soil moisture) for both electrical and gas monitoring was highlighted. Indeed, we observed highest conductivities of the ground at the end of summer and during fall, seasons when the piezometer probes measured the warmer groundwater temperatures (Figure 9(b)). The inverted resistivity sections were adjusted from temperature effect [52,53], with a correction limited at groundwater level. In piezometer Pz2, at 5.29 m depth, the highest variation of temperature over a year is of the order of 2 ∘ C (Figure 9(b)), which corresponds to a variation of resistivity (1.4% per ∘ C) lower than 3%. Nonetheless, a significant decrease in resistivity up to 15% was observed in the area of biodegradation at the end of the summer of 2014 and such a seasonal variation cannot be seen on water conductivity (Figure 9(a)). However, there is also a seasonal variation of the bacterial activity: when groundwaters are warmer, bacterial activity is more intensive and degrades more hydrocarbons and should produce more conductive metabolites [58][59][60][61][62]. Similarly, gas analyses, mainly CO 2 fluxes measurements, showed a seasonal variation. However, this should be mainly due to variation of soil respiration processes; otherwise we would have observed a seasonal variation in BTEX isotopic signatures. On the basis of this analysis, it is possible to make an assertion that seasonal variation of biodegradation processes was detected by our nonintrusive monitoring tools, especially by geoelectrical measurements. However, the permeability of the ground is low and the groundwater table velocity is around 17 m/year. The oxygen from the barrier diffused slowly and its effect was not detected by both direct and indirect monitoring. The seasonal variation of bacterial activity was not yet "erased" by the continuous oxygen injection. Nonetheless, our first results over a year of monitoring show the interest of using geophysical methods and gas analyses to monitor and evaluate in situ remediation. By applying the technology described here, fewer piezometers will be required for the monitoring and the understanding of bioremediation processes, leading to significant cost saving due to fewer monitoring requirements (piezometers, samples, and lab analyses) and more optimized remedial applications based on rapid identification of missed target zones.